nur-dev commited on
Commit
fe31b61
·
verified ·
1 Parent(s): 8094c0c

Publish reproducibility artifacts (part 2)

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. models/temperature/n64_s20852/factorized_ridge.json +1 -0
  2. models/temperature/n64_s20852/factorized_ridge.safetensors +3 -0
  3. models/temperature/n64_s20852/joint_gaussian.json +1 -0
  4. models/temperature/n64_s20852/joint_gaussian.safetensors +3 -0
  5. models/temperature/n64_s20852/joint_neural.json +1 -0
  6. models/temperature/n64_s20852/joint_neural.safetensors +3 -0
  7. models/temperature/n64_s20853/additive_diag.json +1 -0
  8. models/temperature/n64_s20853/additive_diag.safetensors +3 -0
  9. models/temperature/n64_s20853/additive_full.json +1 -0
  10. models/temperature/n64_s20853/additive_full.safetensors +3 -0
  11. models/temperature/n64_s20853/factorized_classifier.json +1 -0
  12. models/temperature/n64_s20853/factorized_classifier.safetensors +3 -0
  13. models/temperature/n64_s20853/factorized_kernel.json +1 -0
  14. models/temperature/n64_s20853/factorized_kernel.safetensors +3 -0
  15. models/temperature/n64_s20853/factorized_lda.json +1 -0
  16. models/temperature/n64_s20853/factorized_lda.safetensors +3 -0
  17. models/temperature/n64_s20853/factorized_logistic.json +1 -0
  18. models/temperature/n64_s20853/factorized_logistic.safetensors +3 -0
  19. models/temperature/n64_s20853/factorized_nb.json +1 -0
  20. models/temperature/n64_s20853/factorized_nb.safetensors +3 -0
  21. models/temperature/n64_s20853/factorized_neural.json +1 -0
  22. models/temperature/n64_s20853/factorized_neural.safetensors +3 -0
  23. models/temperature/n64_s20853/factorized_ridge.json +1 -0
  24. models/temperature/n64_s20853/factorized_ridge.safetensors +3 -0
  25. models/temperature/n64_s20853/joint_gaussian.json +1 -0
  26. models/temperature/n64_s20853/joint_gaussian.safetensors +3 -0
  27. models/temperature/n64_s20853/joint_neural.json +1 -0
  28. models/temperature/n64_s20853/joint_neural.safetensors +3 -0
  29. paper/build_figures.py +69 -0
  30. paper/sensitivity_analysis.py +147 -0
  31. protocols/day_transfer_protocol_01.json +61 -0
  32. protocols/decision_sufficiency_protocol_01.json +86 -0
  33. protocols/external_sufficiency_protocol_01.json +24 -0
  34. protocols/factorial_transfer_protocol_01.json +244 -0
  35. provenance/controlled/compute_ledger.json +64 -0
  36. provenance/controlled/confirmation_benchmark.json +0 -0
  37. provenance/controlled/preconfirmation_lock.json +22 -0
  38. provenance/external/diagnostic_review.json +0 -0
  39. provenance/external/preconfirmation_lock.json +17 -0
  40. provenance/external/sources/manifest.json +213 -0
  41. provenance/external/sources/source_definitions.json +40 -0
  42. provenance/external/task_manifest.json +2608 -0
  43. provenance/hydraulic/confirmation_lock.json +23 -0
  44. provenance/hydraulic/opened_hydraulic_summary.json +695 -0
  45. provenance/temperature/acquisition_summary.json +678 -0
  46. provenance/temperature/cost_and_replay.json +1006 -0
  47. provenance/temperature/forecast_audit.json +1809 -0
  48. provenance/temperature/incomplete_cohort_evaluation.py +96 -0
  49. provenance/temperature/incomplete_cohort_evaluation_lock.json +7 -0
  50. provenance/temperature/lock.json +24 -0
models/temperature/n64_s20852/factorized_ridge.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["parts",{"list":[{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["count",{"tensor":"t000002"}],["centers",{"tensor":"t000003"}],["labels",{"tensor":"t000004"}],["weights",{"tensor":"t000005"}],["size",10]]},{"dict":[["mean",{"tensor":"t000006"}],["cov",{"tensor":"t000007"}],["count",{"tensor":"t000008"}],["centers",{"tensor":"t000009"}],["labels",{"tensor":"t000010"}],["weights",{"tensor":"t000011"}],["size",3]]}]}],["sizes",{"tuple":[10,3]}],["kind","factorized_ridge"],["grid",{"tensor":"t000012"}],["neural_parameters",0]]}],["mean",{"tensor":"t000013"}],["scale",{"tensor":"t000014"}],["fit_ids",{"tensor":"t000015"}]]}
models/temperature/n64_s20852/factorized_ridge.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a29756a84c52bfdf8eed968219e245d559bb5aff14695ddf3b18c5de68da581
3
+ size 123528
models/temperature/n64_s20852/joint_gaussian.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","joint_gaussian"],["estimated_mean_scalars",1680],["checkpoint",{"dict":[["mean",{"tensor":"t000004"}]]}],["adaptation_fits",1],["neural_parameters",0]]}],["mean",{"tensor":"t000005"}],["scale",{"tensor":"t000006"}],["fit_ids",{"tensor":"t000007"}]]}
models/temperature/n64_s20852/joint_gaussian.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ff13fb733a62fc3c7f7d81d829fa081aae1dc1d7392f68bfa79070b1f7958ae
3
+ size 54776
models/temperature/n64_s20852/joint_neural.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","joint_neural"],["estimated_mean_scalars",19979],["checkpoint",{"dict":[["net.0.weight",{"tensor":"t000004"}],["net.0.bias",{"tensor":"t000005"}],["net.2.weight",{"tensor":"t000006"}],["net.2.bias",{"tensor":"t000007"}]]}],["adaptation_fits",5],["neural_parameters",19979]]}],["mean",{"tensor":"t000008"}],["scale",{"tensor":"t000009"}],["fit_ids",{"tensor":"t000010"}]]}
models/temperature/n64_s20852/joint_neural.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:310d85754172e3e0abc90db4b4b8299becf6de3df2a87b38d546eab16c01ed7d
3
+ size 121468
models/temperature/n64_s20853/additive_diag.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","additive_diag"],["estimated_mean_scalars",784],["checkpoint",{"dict":[["weights",{"tensor":"t000004"}]]}],["adaptation_fits",1],["neural_parameters",0]]}],["mean",{"tensor":"t000005"}],["scale",{"tensor":"t000006"}],["fit_ids",{"tensor":"t000007"}]]}
models/temperature/n64_s20853/additive_diag.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6341c182713477dacd0e5bd1c8b563694aba3d88b72a1248ca1303327547a1a6
3
+ size 47608
models/temperature/n64_s20853/additive_full.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","additive_full"],["estimated_mean_scalars",784],["checkpoint",{"dict":[["weights",{"tensor":"t000004"}]]}],["adaptation_fits",1],["neural_parameters",0]]}],["mean",{"tensor":"t000005"}],["scale",{"tensor":"t000006"}],["fit_ids",{"tensor":"t000007"}]]}
models/temperature/n64_s20853/additive_full.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:192e2336937d519074711d076ffe113a6034417e53f729485f382d19827772c5
3
+ size 47608
models/temperature/n64_s20853/factorized_classifier.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["kind","factorized_classifier"],["sizes",{"tuple":[10,3]}],["state_dict",{"dict":[["0.weight",{"tensor":"t000000"}],["0.bias",{"tensor":"t000001"}],["2.weight",{"tensor":"t000002"}],["2.bias",{"tensor":"t000003"}]]}],["nonlinear",true],["input_dimensions",56],["grid",{"tensor":"t000004"}],["temperatures",{"tensor":"t000005"}],["neural_parameters",19965],["adaptation_fits",5]]}],["mean",{"tensor":"t000006"}],["scale",{"tensor":"t000007"}],["fit_ids",{"tensor":"t000008"}]]}
models/temperature/n64_s20853/factorized_classifier.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a792fc43c751833020e414805d6624b07818a212a712f1fbd284a56e431bb2a5
3
+ size 82500
models/temperature/n64_s20853/factorized_kernel.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["parts",{"list":[{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["count",{"tensor":"t000002"}],["centers",{"tensor":"t000003"}],["labels",{"tensor":"t000004"}],["weights",null],["size",10]]},{"dict":[["mean",{"tensor":"t000005"}],["cov",{"tensor":"t000006"}],["count",{"tensor":"t000007"}],["centers",{"tensor":"t000008"}],["labels",{"tensor":"t000009"}],["weights",null],["size",3]]}]}],["sizes",{"tuple":[10,3]}],["kind","factorized_kernel"],["grid",{"tensor":"t000010"}],["neural_parameters",0]]}],["mean",{"tensor":"t000011"}],["scale",{"tensor":"t000012"}],["fit_ids",{"tensor":"t000013"}]]}
models/temperature/n64_s20853/factorized_kernel.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45c483e1f649dba522385f71820238b6c795c92733a6c8f8f5cf689edbbdc513
3
+ size 117456
models/temperature/n64_s20853/factorized_lda.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["parts",{"list":[{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["count",{"tensor":"t000002"}],["centers",{"tensor":"t000003"}],["labels",{"tensor":"t000004"}],["weights",null],["size",10]]},{"dict":[["mean",{"tensor":"t000005"}],["cov",{"tensor":"t000006"}],["count",{"tensor":"t000007"}],["centers",{"tensor":"t000008"}],["labels",{"tensor":"t000009"}],["weights",null],["size",3]]}]}],["sizes",{"tuple":[10,3]}],["kind","factorized_lda"],["grid",{"tensor":"t000010"}],["neural_parameters",0]]}],["mean",{"tensor":"t000011"}],["scale",{"tensor":"t000012"}],["fit_ids",{"tensor":"t000013"}]]}
models/temperature/n64_s20853/factorized_lda.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:887920532c904a603b5302efabf221c651600c054732f20cf2432c0179dbf627
3
+ size 393448
models/temperature/n64_s20853/factorized_logistic.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["kind","factorized_logistic"],["sizes",{"tuple":[10,3]}],["state_dict",{"dict":[["weight",{"tensor":"t000000"}],["bias",{"tensor":"t000001"}]]}],["nonlinear",false],["input_dimensions",56],["grid",{"tensor":"t000002"}],["temperatures",{"tensor":"t000003"}],["neural_parameters",743],["adaptation_fits",5]]}],["mean",{"tensor":"t000004"}],["scale",{"tensor":"t000005"}],["fit_ids",{"tensor":"t000006"}]]}
models/temperature/n64_s20853/factorized_logistic.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff5a5cd4193d0ee8fb50d3bd1338f4894aaf9fdb22ce7053c040d81b88ab658c
3
+ size 5468
models/temperature/n64_s20853/factorized_nb.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["parts",{"list":[{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["count",{"tensor":"t000002"}],["centers",{"tensor":"t000003"}],["labels",{"tensor":"t000004"}],["weights",null],["size",10]]},{"dict":[["mean",{"tensor":"t000005"}],["cov",{"tensor":"t000006"}],["count",{"tensor":"t000007"}],["centers",{"tensor":"t000008"}],["labels",{"tensor":"t000009"}],["weights",null],["size",3]]}]}],["sizes",{"tuple":[10,3]}],["kind","factorized_nb"],["grid",{"tensor":"t000010"}],["neural_parameters",0]]}],["mean",{"tensor":"t000011"}],["scale",{"tensor":"t000012"}],["fit_ids",{"tensor":"t000013"}]]}
models/temperature/n64_s20853/factorized_nb.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e70aad49f9d2633a156d84a5e4cba9a342473851b84c4bfed332c652ef01d6a0
3
+ size 393448
models/temperature/n64_s20853/factorized_neural.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","factorized_neural"],["estimated_mean_scalars",19936],["checkpoint",{"dict":[["net.0.weight",{"tensor":"t000004"}],["net.0.bias",{"tensor":"t000005"}],["net.2.weight",{"tensor":"t000006"}],["net.2.bias",{"tensor":"t000007"}]]}],["adaptation_fits",5],["neural_parameters",19936]]}],["mean",{"tensor":"t000008"}],["scale",{"tensor":"t000009"}],["fit_ids",{"tensor":"t000010"}]]}
models/temperature/n64_s20853/factorized_neural.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5856558ce7588273bdc722927d7fadb3fe2692e4581bde0ae3a389142ef2dd3
3
+ size 121296
models/temperature/n64_s20853/factorized_ridge.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["parts",{"list":[{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["count",{"tensor":"t000002"}],["centers",{"tensor":"t000003"}],["labels",{"tensor":"t000004"}],["weights",{"tensor":"t000005"}],["size",10]]},{"dict":[["mean",{"tensor":"t000006"}],["cov",{"tensor":"t000007"}],["count",{"tensor":"t000008"}],["centers",{"tensor":"t000009"}],["labels",{"tensor":"t000010"}],["weights",{"tensor":"t000011"}],["size",3]]}]}],["sizes",{"tuple":[10,3]}],["kind","factorized_ridge"],["grid",{"tensor":"t000012"}],["neural_parameters",0]]}],["mean",{"tensor":"t000013"}],["scale",{"tensor":"t000014"}],["fit_ids",{"tensor":"t000015"}]]}
models/temperature/n64_s20853/factorized_ridge.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4abc329811273e561323245bd5bab0bc21f14b11dcebce130f905388bc9c6f4
3
+ size 123528
models/temperature/n64_s20853/joint_gaussian.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","joint_gaussian"],["estimated_mean_scalars",1680],["checkpoint",{"dict":[["mean",{"tensor":"t000004"}]]}],["adaptation_fits",1],["neural_parameters",0]]}],["mean",{"tensor":"t000005"}],["scale",{"tensor":"t000006"}],["fit_ids",{"tensor":"t000007"}]]}
models/temperature/n64_s20853/joint_gaussian.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61a2a8634c41b022d1592c023619ba2eaca9c479b10084336f96e81e24082336
3
+ size 54776
models/temperature/n64_s20853/joint_neural.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"dict":[["model",{"dict":[["mean",{"tensor":"t000000"}],["cov",{"tensor":"t000001"}],["prior",{"tensor":"t000002"}],["hypotheses",{"tensor":"t000003"}],["sizes",{"tuple":[10,3]}],["kind","joint_neural"],["estimated_mean_scalars",19979],["checkpoint",{"dict":[["net.0.weight",{"tensor":"t000004"}],["net.0.bias",{"tensor":"t000005"}],["net.2.weight",{"tensor":"t000006"}],["net.2.bias",{"tensor":"t000007"}]]}],["adaptation_fits",5],["neural_parameters",19979]]}],["mean",{"tensor":"t000008"}],["scale",{"tensor":"t000009"}],["fit_ids",{"tensor":"t000010"}]]}
models/temperature/n64_s20853/joint_neural.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7f1d2465706477e65bdd03115813abd8504589d3cbe749ebcd7c4188c363ae1
3
+ size 121468
paper/build_figures.py ADDED
@@ -0,0 +1,69 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Generate every plotted value and table body from archived result records."""
2
+ from pathlib import Path
3
+ import json,csv,statistics,shutil
4
+ import matplotlib
5
+ matplotlib.use('Agg')
6
+ import matplotlib.pyplot as plt
7
+ import numpy as np
8
+ R=Path(__file__).resolve().parents[1];A=R/'paper';F=A/'figures';T=A/'tables';F.mkdir(exist_ok=True);T.mkdir(exist_ok=True);(A/'metadata').mkdir(exist_ok=True)
9
+ plt.rcParams.update({'font.family':'DejaVu Sans','font.size':11.5,'axes.spines.top':False,'axes.spines.right':False,'pdf.fonttype':42,'savefig.bbox':'tight'})
10
+ NAMES={'additive_full':'Additive Gaussian','additive_diag':'Diagonal Gaussian','factorized_neural':'Factorized forward MLP','joint_neural':'Joint-state forward MLP','factorized_nb':'Factorized Gaussian NB','factorized_lda':'Factorized Gaussian LDA','factorized_kernel':'Factorized kernel','factorized_ridge':'Factorized ridge','joint_gaussian':'Joint-state Gaussian','factorized_logistic':'Factorized logistic','factorized_classifier':'Direct factorized MLP'}
11
+ METHODS={'query_joint':'Joint predictive clustering','posterior_clusters':'Posterior clustering','posterior_bins':'Posterior bins','future_marginal':'Observation clustering','raw_information':'Three features by mutual information','raw_entropy':'Three features by entropy','hash_code':'Fixed hash','identity':'Uncompressed mixture'}
12
+ load=lambda p:json.loads((R/p).read_text())
13
+ control=load('results/controlled/decision.json');day=load('provenance/verification/independent_statistics.json');hydro=load('provenance/hydraulic/opened_hydraulic_summary.json');summary=load('provenance/temperature/summary.json')
14
+ primary={r['method']:r for r in control['primary_metrics']}
15
+ with (T/'controlled.tex').open('w') as f:
16
+ for m in ['query_joint','raw_information','raw_entropy','posterior_clusters','posterior_bins','future_marginal','hash_code','identity']:
17
+ r=primary[m];f.write(f"{METHODS[m]} & {r['utility']:.4f} & {r['representation_loss']:.4f} & {r['model_loss']:.4f} " + chr(92)*2 + chr(10))
18
+ front=list(csv.DictReader((R/'results/controlled/frontier.csv').open()))
19
+ fig,ax=plt.subplots(1,2,figsize=(8,3.7))
20
+ for m,c in [('query_joint','#176b91'),('raw_information','#ba682e'),('posterior_clusters','#767676'),('identity','#277a50')]:
21
+ xs=[64,256,1024,4096];ys=[]
22
+ for n in xs:
23
+ rows=[r for r in front if r['availability']=='full' and r['model']=='support_mixture' and r['method']==m and int(r['label_budget'])==n and int(r['query_budget'])>0]
24
+ assert len(rows)==2,(m,n,len(rows));ys.append(statistics.mean(float(r['utility']) for r in rows))
25
+ ax[0].plot(xs,ys,'o-',color=c,label=METHODS[m])
26
+ ax[0].set_xscale('log',base=2);ax[0].set_xticks(xs,[str(x) for x in xs]);ax[0].set_xlabel('Estimation records');ax[0].set_ylabel('Expected utility');ax[0].set_title('(a) Sample dependence');ax[0].legend(fontsize=9,loc='lower right')
27
+ fam=list(csv.DictReader((R/'results/controlled/primary_families.csv').open()));families=sorted(set(r['family'] for r in fam))
28
+ for m,dx,c in [('query_joint',-.24,'#176b91'),('raw_information',0,'#ba682e'),('posterior_clusters',.24,'#767676')]:
29
+ ys=[float(next(r['utility'] for r in fam if r['family']==fa and r['method']==m)) for fa in families]
30
+ ax[1].bar(np.arange(6)+dx,ys,.23,color=c,label=METHODS[m])
31
+ ax[1].set_xticks(np.arange(6),[x.replace('_','\n') for x in families],rotation=30,ha='right',fontsize=9);ax[1].set_ylabel('Expected utility');ax[1].set_title('(b) All six declared mechanism families');fig.tight_layout();fig.savefig(F/'controlled.pdf');plt.close(fig)
32
+ # Matrix uses independent verification's equal-day scores, never the pooled-episode scores.
33
+ def dscore(model,n):return next(r['top3_equal_day'] for r in day['summary'] if r['model']==model and r['samples']==n)
34
+ def hscore(model,n):
35
+ rows=[r for r in hydro['rows'] if r.get('model')==model and r.get('samples')==n]
36
+ assert len(rows)==1,(model,n,rows);return rows[0]['top3']
37
+ with (T/'all_models.tex').open('w') as f:
38
+ for m in NAMES:
39
+ vals=[hscore(m,64),hscore(m,128),dscore(m,64),dscore(m,128)]
40
+ f.write(NAMES[m]+' & '+' & '.join(f'{v*100:.2f}' for v in vals)+' '+chr(92)*2+chr(10))
41
+ fig,axes=plt.subplots(1,2,figsize=(8,5.2),sharey=True)
42
+ for axis,n in zip(axes,[64,128]):
43
+ y=np.arange(len(NAMES));axis.barh(y-.16,[100*hscore(m,n) for m in NAMES],.3,label='Hydraulic: opened development',color='#ba682e');axis.barh(y+.16,[100*dscore(m,n) for m in NAMES],.3,label='Gas array: four held days',color='#176b91');axis.set_yticks(y,list(NAMES.values()),fontsize=11);axis.set_ylim(len(NAMES)-.5,-.5);axis.set_xlim(0,105);axis.set_xlabel('Unseen-combination Top-3 (%)');axis.set_title(f'{n} total fitting examples');axis.grid(axis='x',alpha=.2)
44
+ axes[1].legend(fontsize=9,loc='lower right');fig.tight_layout();fig.savefig(F/'models.pdf');plt.close(fig)
45
+ fig,axes=plt.subplots(1,3,figsize=(8,3))
46
+ for axis,reference in zip(axes,['factorized_classifier','additive_full','factorized_lda']):
47
+ a=next(r['per_day'] for r in day['summary'] if r['model']=='factorized_neural' and r['samples']==64);b=next(r['per_day'] for r in day['summary'] if r['model']==reference and r['samples']==64);g=100*(np.array(a)-np.array(b));axis.axhline(0,color='black',lw=.8);axis.scatter(range(1,5),g,color='#176b91',s=45);axis.set_xticks(range(1,5));axis.set_ylim(-20,50);axis.set_xlabel('Held day (ordered)');axis.set_title(NAMES[reference],fontsize=10);axis.grid(alpha=.15)
48
+ axes[0].set_ylabel('Forward MLP gain (percentage points)');fig.tight_layout();fig.savefig(F/'days.pdf');plt.close(fig)
49
+ with (T/'seen_models.tex').open('w') as f:
50
+ for m in NAMES:
51
+ vals=[]
52
+ for n in [64,128]:
53
+ rows=[r for r in summary['rows'] if r.get('model')==m and r.get('samples')==n and r.get('stratum')=='seen' and r.get('day') is not None]
54
+ assert len(rows)==4;vals.append(statistics.mean(r['top3'] for r in rows))
55
+ f.write(NAMES[m]+' & '+' & '.join(f'{100*v:.2f}' for v in vals)+' '+chr(92)*2+chr(10))
56
+ proof={'controlled_gain':control['compact_comparisons'][3], 'temperature_scores':{m:dscore(m,64) for m in NAMES},'hydraulic_scores':{m:hscore(m,64) for m in NAMES},'figure_sources':['results/controlled/frontier.csv','results/controlled/primary_families.csv','provenance/verification/independent_statistics.json','provenance/hydraulic/opened_hydraulic_summary.json']}
57
+ (A/'metadata/figure_values.json').write_text(json.dumps(proof,indent=2)+'\n')
58
+ diagnostics=load('results/sensitivity_analysis/frozen_diagnostics.json')
59
+ with (T/'sensitivity.tex').open('w') as f:
60
+ for cohort,label in [('hydraulic','Hydraulic'),('temperature','Gas array')]:
61
+ for n in [64,128]:
62
+ rows={r['model']:r for r in diagnostics['sensor']['summary'] if r['cohort']==cohort and r['samples']==n}
63
+ values=[label,str(n),str(rows['factorized_neural']['observed_combinations'])]
64
+ values += [f"{100*rows[m]['macro_combination_top3']:.2f}" for m in ['factorized_neural','factorized_classifier','additive_full','factorized_lda']]
65
+ f.write(' & '.join(values)+' '+chr(92)*2+chr(10))
66
+ print('Generated 3 vector figures and 4 table bodies')
67
+
68
+ for filename,macro in [("controlled.tex","ControlledRows"),("all_models.tex","AllModelRows"),("seen_models.tex","SeenModelRows"),("sensitivity.tex","SensitivityRows")]:
69
+ p=T/filename;p.write_text(chr(92)+"newcommand{"+chr(92)+macro+"}{%"+chr(10)+p.read_text()+"}%"+chr(10))
paper/sensitivity_analysis.py ADDED
@@ -0,0 +1,147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Post-evaluation checks of frozen models; no fitting or recipe selection."""
2
+ from pathlib import Path
3
+ import argparse,hashlib,importlib.util,itertools,json,sys,time
4
+ PROCESS_START=time.perf_counter()
5
+ import torch
6
+
7
+ def aggregate(value,days):
8
+ if days is None:return float(value.mean())
9
+ return float(torch.stack([value[days==d].mean() for d in days.unique()]).mean())
10
+
11
+ def macro_combination(value,codes,days):
12
+ if days is None:days=torch.zeros_like(codes)
13
+ return float(torch.stack([torch.stack([value[(days==d)&(codes==h)].mean()
14
+ for h in codes[days==d].unique()]).mean() for d in days.unique()]).mean())
15
+
16
+ def sensor_checks(replay,device):
17
+ output=[];supports=[];maximum=0.;comparisons=[]
18
+ with torch.no_grad():
19
+ for cohort,sizes in [('hydraulic',[3,4,3,4]),('temperature',[10,3])]:
20
+ p=replay.load(f'data/{cohort}/prepared',device)
21
+ if cohort=='hydraulic':p.update(p['masks'])
22
+ records=replay.load(f'results/{cohort}/predictions',device)['results']
23
+ for r in records:
24
+ n,seed,name=r['samples'],r['seed'],r['model']
25
+ checkpoint=replay.load(f'models/{cohort}/n{n}_s{seed}/{name}',device)
26
+ fit=checkpoint['fit_ids'];held=r['held_rows'];mask=r['unseen']
27
+ assert bool(p['source'][fit].all()) and not bool(p['unseen'][fit].any())
28
+ prob=replay.predict(checkpoint['model'],(p['x'][held]-checkpoint['mean'])/checkpoint['scale'])
29
+ truth=replay.factorial.joint_index(p['h'][held],sizes)
30
+ scores={f'top{k}':replay.factorial.topk_credit(prob,truth,k) for k in (1,3)}
31
+ scores.update(deterministic_top1=(prob.argmax(-1)==truth).double(),
32
+ deterministic_top3=(prob.topk(3,-1).indices==truth[:,None]).any(-1).double())
33
+ scores.update(nll=-prob.gather(1,truth[:,None]).squeeze(-1).clamp_min(1e-300).log(),
34
+ confidence=prob.max(-1).values,top3_mass=prob.topk(3,-1).values.sum(-1))
35
+ for key,value in scores.items():
36
+ error=float((value-r['metrics'][key]).abs().max());maximum=max(maximum,error)
37
+ assert error<1e-8,(cohort,n,seed,name,key,error)
38
+ codes=truth[mask];days=r['days'][mask] if cohort=='temperature' else None
39
+ row=dict(cohort=cohort,samples=n,seed=seed,model=name,records=int(mask.sum()),
40
+ observed_combinations=len(codes.unique()),
41
+ primary_top1=aggregate(scores['top1'][mask],days),
42
+ primary_top3=aggregate(scores['top3'][mask],days),
43
+ deterministic_top3=aggregate(scores['deterministic_top3'][mask],days),
44
+ macro_combination_top3=macro_combination(scores['top3'][mask],codes,days),
45
+ top3_probability_mass=aggregate(scores['top3_mass'][mask],days))
46
+ distance=(p['h'][held][mask,None]!=p['h'][fit][None]).sum(-1).min(-1).values
47
+ assert bool((distance>=1).all())
48
+ row['nearest_fitting_combination']=[dict(hamming_distance=int(d),records=int((distance==d).sum()),
49
+ top3=aggregate(scores['top3'][mask][distance==d],days[distance==d] if days is not None else None))
50
+ for d in distance.unique()]
51
+ top4=prob[mask].topk(4,-1).values.clamp_min(1e-300)
52
+ row['median_third_fourth_log_probability_margin']=float((top4[:,2].log()-top4[:,3].log()).median())
53
+ row['per_combination']=[dict(code=int(h),levels=p['h'][held][mask][(codes==h).nonzero()[0,0]].tolist(),
54
+ records=int((codes==h).sum()),top3=aggregate(scores['top3'][mask][codes==h],days[codes==h] if days is not None else None))
55
+ for h in codes.unique()]
56
+ output.append(row)
57
+ if name=='factorized_neural':
58
+ supports.append(dict(cohort=cohort,samples=n,seed=seed,
59
+ factor_level_counts=[torch.bincount(p['h'][fit,i],minlength=k).tolist() for i,k in enumerate(sizes)],
60
+ source_combinations=len(replay.factorial.joint_index(p['h'][fit],sizes).unique())))
61
+ for n in [64,128]:
62
+ for control in ['factorized_classifier','additive_full','factorized_lda']:
63
+ a={r['seed']:r for r in output if r['cohort']==cohort and r['samples']==n and r['model']=='factorized_neural'}
64
+ b={r['seed']:r for r in output if r['cohort']==cohort and r['samples']==n and r['model']==control}
65
+ comparisons.append(dict(cohort=cohort,samples=n,control=control,
66
+ paired_seed_gains=[dict(seed=s,primary=a[s]['primary_top3']-b[s]['primary_top3'],
67
+ macro_combination=a[s]['macro_combination_top3']-b[s]['macro_combination_top3']) for s in sorted(a)]))
68
+ summary=[]
69
+ for cohort in ['hydraulic','temperature']:
70
+ for n in [64,128]:
71
+ for name in sorted({r['model'] for r in output}):
72
+ rows=[r for r in output if r['cohort']==cohort and r['samples']==n and r['model']==name]
73
+ summary.append(dict(cohort=cohort,samples=n,model=name,observed_combinations=rows[0]['observed_combinations'],
74
+ primary_top3=sum(r['primary_top3'] for r in rows)/len(rows),
75
+ deterministic_top3=sum(r['deterministic_top3'] for r in rows)/len(rows),
76
+ macro_combination_top3=sum(r['macro_combination_top3'] for r in rows)/len(rows),
77
+ seed_top3_min=min(r['primary_top3'] for r in rows),seed_top3_max=max(r['primary_top3'] for r in rows),
78
+ mean_top3_probability_mass=sum(r['top3_probability_mass'] for r in rows)/len(rows)))
79
+ assert len(output)==132
80
+ return dict(checkpoints=len(output),maximum_replay_error=maximum,summary=summary,
81
+ seed_comparisons=comparisons,fitting_support=supports,individual_fits=output)
82
+
83
+ def enumerate_trees(device,budget):
84
+ """All depth-limited binary-query trees, independently of the DP recurrence."""
85
+ outcomes=list(itertools.product([0,1],repeat=2));terminal=[[1.,-2.],[-2.,1.],[0.,0.]]
86
+ trees=[([a]*4,[0]*4) for a in range(3)]
87
+ for first in [0,1]:
88
+ other=1-first
89
+ branch_options=[('stop',a,a) for a in range(3)]
90
+ if budget==2:branch_options += [('query',a,b) for a in range(3) for b in range(3)]
91
+ for branches in itertools.product(branch_options,repeat=2):
92
+ actions=[];queries=[]
93
+ for outcome in outcomes:
94
+ kind,a,b=branches[outcome[first]]
95
+ actions.append(a if kind=='stop' or outcome[other]==0 else b)
96
+ queries.append(1+int(kind=='query'))
97
+ trees.append((actions,queries))
98
+ assert len(trees)==(21 if budget==1 else 291)
99
+ utilities=torch.tensor([[[terminal[a][h] for a in actions] for h in range(2)] for actions,_ in trees],device=device,dtype=torch.float64)
100
+ queries=torch.tensor([q for _,q in trees],device=device,dtype=torch.float64)[:,None].expand(-1,2,-1)
101
+ return utilities.flatten(1),queries.flatten(1)
102
+
103
+ def planner_check(replay,device):
104
+ core=replay.core;worlds=replay.load('data/controlled/worlds',device);maximum=0.;comparisons=0
105
+ names=['identity','query_joint','posterior_clusters','posterior_bins','future_marginal','raw_information','raw_entropy','hash_code']
106
+ trees={b:enumerate_trees(device,b) for b in [1,2]}
107
+ with torch.no_grad():
108
+ for stream in range(3):
109
+ fitted=core.estimate(core.count_samples(worlds['samples'][:,stream,:1024],6),'support_mixture')
110
+ for name in names:
111
+ code,k=core.encode(fitted,name);law=core.push(fitted,code,k);policy=core.plan(law)
112
+ flat=law.flatten(2)
113
+ for budget in [1,2]:
114
+ reward,queries=trees[budget]
115
+ for ci,cost in enumerate([0.,.01,.05]):
116
+ expected=(flat@(reward-cost*queries).T).max(-1).values.sum(-1)
117
+ error=float((expected-policy['expected'][:,budget,ci]).abs().max())
118
+ maximum=max(maximum,error);comparisons+=len(law)
119
+ assert error<1e-12,(name,budget,cost,error)
120
+ return dict(worlds=192,streams=3,encodings=len(names),utility_comparisons=comparisons,
121
+ depth_one_trees=21,depth_two_trees=291,maximum_absolute_error=maximum,
122
+ verification='All deterministic binary-query trees; separate from the dynamic-programming recurrence.')
123
+
124
+ def main():
125
+ parser=argparse.ArgumentParser(description=__doc__);parser.add_argument('--release',type=Path)
126
+ parser.add_argument('--device',default='cuda:1');parser.add_argument('--output',type=Path,required=True);a=parser.parse_args()
127
+ root=a.release or Path(__file__).resolve().parents[1]
128
+ if not (root/'reproduce.py').is_file():root=root/'release/huggingface'
129
+ root=root.resolve();sys.path.insert(0,str(root))
130
+ spec=importlib.util.spec_from_file_location('frozen_replay',root/'reproduce.py')
131
+ replay=importlib.util.module_from_spec(spec);spec.loader.exec_module(replay)
132
+ assert torch.cuda.is_available() and a.device.startswith('cuda')
133
+ t=time.perf_counter();torch.cuda.set_device(a.device);torch.cuda.init();torch.cuda.synchronize()
134
+ init=time.perf_counter()-t;torch.backends.cuda.matmul.allow_tf32=False
135
+ t=time.perf_counter();sensor=sensor_checks(replay,a.device);torch.cuda.synchronize();sensor_seconds=time.perf_counter()-t
136
+ t=time.perf_counter();planner=planner_check(replay,a.device);torch.cuda.synchronize();planner_seconds=time.perf_counter()-t
137
+ model_hashes={str(p.relative_to(root)):hashlib.sha256(p.read_bytes()).hexdigest() for p in sorted((root/'models').rglob('*')) if p.is_file()}
138
+ report=dict(post_evaluation_descriptive_analysis=True,new_training=False,new_thresholds=False,new_confirmation=False,
139
+ sensor=sensor,enumerative_planner_check=planner,model_files_sha256=model_hashes,
140
+ compute=dict(cuda_initialization_seconds=init,sensor_diagnostics_seconds=sensor_seconds,
141
+ planner_verification_seconds=planner_seconds,process_seconds_after_standard_library_import=time.perf_counter()-PROCESS_START,
142
+ peak_cuda_bytes=torch.cuda.max_memory_allocated(a.device),device=torch.cuda.get_device_name(a.device)))
143
+ a.output.parent.mkdir(parents=True,exist_ok=True);a.output.write_text(json.dumps(report,indent=2)+'\n')
144
+ print(json.dumps(dict(checkpoints=sensor['checkpoints'],maximum_replay_error=sensor['maximum_replay_error'],
145
+ planner=planner,compute=report['compute']),indent=2))
146
+
147
+ if __name__=='__main__':main()
protocols/day_transfer_protocol_01.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "day_transfer_01",
3
+ "objective": "Independent, separately registered full-hypothesis cross-day retest of unchanged compositional world models. The preceding strict-clock eligibility failure is final and is not amended or replaced.",
4
+ "source": {
5
+ "uci": 487,
6
+ "doi": "10.24432/C5S302",
7
+ "author": "Javier Burgues, Institute of Bioengineering of Catalonia",
8
+ "url": "https://archive.ics.uci.edu/static/public/487/gas+sensor+array+temperature+modulation.zip",
9
+ "license": "CC BY 4.0",
10
+ "access": "Only official metadata and README accessed before lock; archive and numerical values unopened."
11
+ },
12
+ "source_definition": "13 timestamp-named files in order. First nine days fitting pool; last four days held. Preserve every day; any missing or invalid day fails complete cohort. Native nondecreasing seconds and source row order allowed; repeat timestamps never imply extra independent observations. One 900-second provider protocol block is one episode; require >=850s coverage. At least90 episodes per day,128 fitting records,and8 unseen-binding cases per held day.",
13
+ "features": "Four consecutive 225-second means for each of14 resistance channels (columns6:20),56 values. No CO,humidity,temperature,flow or heater reference enters features. Preserve source MOhm. No feature selection.",
14
+ "ontology": "10 author CO levels uniformly0..20ppm, identified by round(9*median_last_quarter_CO/20),clamped0..9; three measured RH bands <=35,(35,55],>55 percent using last-quarter reference mean. Full30 states. Reference labels are supervision only.",
15
+ "split": "Hold first6 of30 ontology bindings by SHA256(20821:factor0,factor1). Entire binding excluded from first9days fitting pool. Report all held observations, seen/unseen strata, each day and unsupported factor levels; no after-opening selection.",
16
+ "sample_budgets": [
17
+ 64,
18
+ 128
19
+ ],
20
+ "seeds": [
21
+ 20851,
22
+ 20852,
23
+ 20853
24
+ ],
25
+ "sampling": "Same random fitting prefix across all11 models and both budgets; no class balancing or target tuning.",
26
+ "models": [
27
+ "additive_full",
28
+ "additive_diag",
29
+ "factorized_neural",
30
+ "joint_neural",
31
+ "factorized_nb",
32
+ "factorized_lda",
33
+ "factorized_kernel",
34
+ "factorized_ridge",
35
+ "joint_gaussian",
36
+ "factorized_logistic",
37
+ "factorized_classifier"
38
+ ],
39
+ "candidates": [
40
+ "additive_full",
41
+ "factorized_neural"
42
+ ],
43
+ "algorithms": "Additive: all factor one-hots plus intercept, ridge 8, intercept ridge .01, analytic leave-one-out residuals. Factorized neural: full factor one-hots -> SiLU hidden -> observation means. Joint neural: one-hot complete binding -> matched SiLU hidden -> observation means, unknown bindings use uninformative source mean. Hidden widths give <=20000 counted trainable parameters and near-matched factorized/nonfactorized budgets. Both neural decoders fit 400 AdamW steps, lr .003, weight_decay .01; four source folds estimate held residuals plus one full fit, all charged. Source residual covariance is .5 full + .5 diagonal with .05 identity floor; diagonal ablation uses only the diagonal. Priors uniform over the supplied full hypothesis ontology, including unseen bindings.",
44
+ "controls": "Independent-factor Gaussian NB and shared-covariance Gaussian, independent-factor whitened kernel with bandwidth .7 and .5 Dirichlet smoothing, independent-factor ridge classifier with ridge 8, joint-binding Gaussian with shrink-to-source fallback. Use all observed coordinates and the same fitting labels. All factorized controls can predict unseen bindings; do not make the comparison only against memorization. Also include class-balanced multinomial logistic regression and a matched <=20000-parameter SiLU direct classifier, both predicting independent factor probabilities, with the same 400-step optimizer and four out-of-fitting-fold source predictions plus full fit. Their four-point source temperature grid (.5,1,2,4) is chosen only on source cross-fit log loss; every fit is charged. All source- and target-state priors remain uniform.",
45
+ "normalization": "Feature mean and standard deviation from the actual 64/128 fitting records only. No full-source label-conditional quantiles. Standard deviation floor 1e-6. All models receive identical normalized numerical evidence.",
46
+ "uncertainty": "20000 paired episode draws independently within each of4held days, equal-day weights,seed20891,family40. Additionally paired day-cluster bootstrap; require both intervals positive. Three seed results averaged as independent fits, never a prediction ensemble. Only4helddays: no population-of-systems generalization.",
47
+ "success": "Primary64examples: >=.10 unseen-binding jointTop3 gain over all7strong baselines,positive simultaneous95% lower bound in both bootstrap analyses,positive point gain on each held day. Overall superiority additionally requires positive intervals against all10alternatives. Neural-specific claim requires same against additive reference.128-label result cannot rescue64primary. All13days and complete cost ledger required.",
48
+ "acquisition": {
49
+ "policies": [
50
+ "value_forced",
51
+ "value_stop",
52
+ "variance_forced",
53
+ "variance_stop",
54
+ "static_forced",
55
+ "random_forced"
56
+ ],
57
+ "cohort": "First8unseenbinding held episodes per held day under SHA256(20901:recordID); all primary records remain.64examples,firstseed,2fixed world models.2/4/8queries,unitcost.01,unchanged exact one-step value operations. Secondary only. No acquisition advantage without>=.10gain over allcheap controls and paired simultaneous family120 bounds; actual processing cost separately reported."
58
+ },
59
+ "compute": "CUDA float64 statistics, probability operations, exact interval decisions and bootstrap; float32 auxiliary neural fitting. CPU archive text parsing, provenance, hashing and report aggregation only. Count data preparation, normalization, every full/cross-fit adaptation, full-hypothesis mean materialization, covariance construction, predictions, query decisions, peak GPU memory, trainable/checkpoint/derived-state bytes. No CPU offload, DeepSpeed, Qwen executions or model collaboration.",
60
+ "stopping": "No new model,capacity,threshold,feature,sample,day or cohort revision after numeric access. Code and inherited modules hashed before source download. Preserve failed original source-clock study and every preceding result. No human-review dependency,no Qwen/backbone execution,no article-only output."
61
+ }
protocols/decision_sufficiency_protocol_01.json ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "Decision sufficiency of estimated predictive representations",
3
+ "version": 1,
4
+ "scope": "Controlled finite-world study, not fresh externally authored model confirmation or operational diagnosis",
5
+ "hypothesis": "At matched eight-symbol capacity, retaining estimated joint hypothesis/future-observation relationships improves adaptive decision value over posterior-only and observation-only compression. Neither exact inference nor current-decision sufficiency repairs missing conditional relationships.",
6
+ "worlds": {
7
+ "development": {"families": ["independent", "context", "parity", "uninformative"], "per_family": 8, "seed": 18401},
8
+ "confirmation": {"families": ["independent", "context", "parity", "uninformative", "redundant", "orientation_alias"], "per_family": 32, "seed": 18501},
9
+ "held_mechanisms": ["redundant", "orientation_alias"],
10
+ "hypotheses": 2,
11
+ "current_binary_measurements": 6,
12
+ "retrievable_binary_measurements": 2,
13
+ "joint_states": 512,
14
+ "availability_conditions": ["full", "three_available"],
15
+ "unavailable_evidence": "Three hidden current channels are never retrospectively queryable. Only the two recorded future channels can be revealed after paying a cost.",
16
+ "freshness": "Generate confirmation joint laws and sample streams only after code, tests, development output and protocol hashes are locked. These are newly generated controlled worlds, not independent external authors or physical evidence. No opened bnRep or diagnostic cases count as fresh confirmation."
17
+ },
18
+ "task": {"actions": ["identify_H0", "identify_H1", "abstain"], "correct_utility": 1.0, "incorrect_utility": -2.0, "abstain_utility": 0.0, "query_costs": [0.0, 0.01, 0.05], "query_budgets": [0, 1, 2], "tie_tolerance": 1e-10},
19
+ "estimation": {
20
+ "label_budgets": [64, 256, 1024, 4096],
21
+ "sample_seeds": [18601, 18602, 18603],
22
+ "nested_samples": true,
23
+ "models": ["joint_jeffreys", "naive_bayes", "support_mixture"],
24
+ "joint_pseudocount_per_cell": 0.5,
25
+ "naive_beta_per_state": 0.5,
26
+ "mixture_weight_joint": "N / (N + 512), fixed before confirmation",
27
+ "primary_model": "support_mixture",
28
+ "target_truth_access": "Estimator and encoder accept only sampled labeled tuples and declared availability. True joint laws enter only sample generation and separate evaluator after policy construction."
29
+ },
30
+ "representations": {
31
+ "symbols": 8,
32
+ "methods": ["identity", "query_joint", "posterior_clusters", "posterior_bins", "future_marginal", "raw_information", "raw_entropy", "hash_code"],
33
+ "query_joint": "Weighted deterministic eight-means of sqrt estimated P(H,Y0,Y1 | E). 32 iterations, farthest-point initialization weighted by estimated P(E).",
34
+ "posterior_clusters": "Same weighted eight-means and iterations on sqrt estimated P(H | E).",
35
+ "posterior_bins": "Eight fixed equal-width posterior bins.",
36
+ "future_marginal": "Same weighted eight-means on sqrt estimated P(Y0,Y1 | E), removing hypothesis relationships.",
37
+ "raw_information": "Three current bits with greatest estimated individual mutual information with H, all eight resulting patterns retained.",
38
+ "raw_entropy": "Three current bits with greatest estimated marginal entropy, all eight patterns retained.",
39
+ "hash_code": "Fixed probability-blind hash of current evidence coordinates to eight codes.",
40
+ "identity": "Retain all available raw patterns; strongest same-estimator information reference, not capacity matched.",
41
+ "code_model": "Push estimated law through encoder and apply the same exact bounded planner. No gold posterior, policy or future outcome enters encoding.",
42
+ "storage": "Count encoding lookup, resident compressed joint, fitted statistical state, temporary fit memory, planner tables, materialization and all construction work. Compression is not free and cannot add information."
43
+ },
44
+ "decomposition": {
45
+ "total_regret": "V(full raw truth) - V(realized estimated policy)",
46
+ "information": "V(full raw truth) - V(available raw truth)",
47
+ "representation": "V(available raw truth) - V(encoded truth)",
48
+ "model": "V(encoded truth) - V(realized policy planned exactly under estimated encoded model)",
49
+ "inference": "Zero for main exact-planning arms; verify against an independently implemented scalar enumerator. A no-abstention action heuristic is a separate negative control.",
50
+ "note": "All values are ex-ante world expected utility over admissible policies, including query costs. Static and adaptive sufficiency are evaluated separately."
51
+ },
52
+ "secondary_adequacy_diagnostic": {
53
+ "conditional_dispersion": "Expected total variation between each raw-state conditional P(H,Y|R) and its encoded-cell conditional P(H,Y|Z). Compute estimated and evaluator-only true dispersion separately.",
54
+ "bound": "Representation loss <= (3 + budget*query_cost) * true conditional dispersion, verified at every evaluated state budget and cost. This follows from the bounded utility range and is not a new theorem.",
55
+ "limitation": "Estimated dispersion is internal model adequacy only; it is not a certificate for the true law when the model is wrong.",
56
+ "registered_before_confirmation": true,
57
+ "primary_thresholds_or_encoders_changed": false
58
+ },
59
+ "primary": {
60
+ "model": "support_mixture", "label_budget": 1024, "availability": "full", "method": "query_joint",
61
+ "aggregation": "Equal weight on both positive query budgets, all three costs, three independent sample streams, worlds within families, then all six families. No selected publication, cost, seed or mechanism subsets.",
62
+ "compact_controls": ["posterior_clusters", "posterior_bins", "future_marginal", "raw_information", "raw_entropy", "hash_code"],
63
+ "mean_utility_gain_at_least": 0.01,
64
+ "all_paired_lower_95_bounds_above": 0.0,
65
+ "bootstrap_draws": 10000, "bootstrap_seed": 18701,
66
+ "resampling": "Worlds paired within each fixed family; costs, masks, budgets, samples and derived observations are not independent cases. Simultaneous Bonferroni bounds across the six eligible compact controls.",
67
+ "full_information_noninferiority_margin": 0.01,
68
+ "note": "A compact-control advantage is not superiority to uncompressed estimation, not evidence for learned control, and not external generalization."
69
+ },
70
+ "cost_promotion": {
71
+ "quality": "Lower paired 95% query_joint minus strongest full-information statistical reference > -0.01; all eligible full models reported",
72
+ "cost": "At least 20% measured complete cost reduction including estimation, encoding, model materialization, planning and repeated requests; resident and transient storage reported separately",
73
+ "repeated_request_counts": [1, 32, 1024, 65536], "timing_repeats": 11,
74
+ "external_confirmation_required_before_practical_promotion": true,
75
+ "no_production_latency_or_engineer_time_claim": true
76
+ },
77
+ "controls": ["Faithful full explicit law", "Faithful unavailable-evidence law", "Empirical estimated laws without supplied probabilities", "Known exact posterior-only adaptive counterexample", "Same exact planner for every representation", "Independent scalar exact planner", "No-abstention inference negative control"],
78
+ "falsification": [
79
+ "If query_joint fails its strongest matched compact controls, reject its registered advantage.",
80
+ "If estimated representations preserve their own modeled decisions but lose true value, attribute loss to model/representation error rather than claiming reasoning transfer.",
81
+ "If missing raw evidence already reduces oracle value, more computation cannot recover that information.",
82
+ "If posterior-only compression is static sufficient but loses adaptive value in the exact counterexample, current posterior sufficiency alone is inadequate for this query problem.",
83
+ "No architecture escalation, new controller training, target threshold adjustment or favorable subgroup promotion after confirmation."
84
+ ],
85
+ "execution": {"device": "cuda:2", "dtype": "float64", "new_neural_modules": 0, "backbone_executions": 0, "controller_refits": 0, "prior_results_immutable": true}
86
+ }
protocols/external_sufficiency_protocol_01.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "external_decision_sufficiency_01",
3
+ "scope": "External recorded-data validation of frozen representation construction and exact bounded planning; sequential feature reveal is authored here, not supplied by the data providers.",
4
+ "device": "cuda:2",
5
+ "dtype": "float64",
6
+ "new_neural_modules": 0,
7
+ "backbone_executions": 0,
8
+ "sources": {
9
+ "hydraulic": {"uci_id":447,"doi":"10.24432/C5CW21","url":"https://archive.ics.uci.edu/static/public/447/condition+monitoring+of+hydraulic+systems.zip","license":"UCI CC BY 4.0","features":["PS1","PS2","PS3","PS4","FS1","TS1","EPS1","VS1"],"aggregation":"arithmetic mean of all samples of each recorded 60-second cycle","tasks":["cooler","valve","pump","accumulator"],"normal":[100,100,0,130],"context_column":[1,0,0,0],"source_context":[100,100,100,100],"target_contexts":[[73,80,90],[3,20],[3,20],[3,20]],"group":"all cycles with identical four physical condition labels, irrespective of stability; SHA256(profile) mod 2 separates adaptation and evaluation"},
10
+ "gas": {"uci_id":224,"doi":"10.24432/C5RP6W","url":"https://archive.ics.uci.edu/static/public/224/gas+sensor+array+drift+dataset.zip","license":"UCI CC BY 4.0; legacy description additionally restricts use to research, which this study respects","features_zero_based":[1,9,17,25,33,41,49,57],"feature_meaning":"normalized resistance response from sensors 1 through 8, in provider order","tasks":"all 15 lexicographic pairs of gas labels 1 through 6","source_batches":[1,2,3],"target_batches":[6,7,8,9,10],"group":"within each batch, contiguous blocks of 16 original rows before pair filtering; SHA256(batch,block) mod 2 separates adaptation and evaluation"}
11
+ },
12
+ "adapter": {"current_bits":6,"query_bits":2,"binarization":"strictly greater than each feature's source-only median; source unlabeled preprocessing uses all source records and is charged","no_target_normalization":true,"split_seed":20101,"sample_seeds":[20111,20112,20113],"samples_without_replacement":true,"source_labels_maximum":1024},
13
+ "eligibility": {"source_minimum":64,"adaptation_minimum_for_primary":64,"evaluation_minimum":32,"evaluation_minimum_per_hypothesis":8,"eligible_groups_minimum":4,"exclude_only_for_registered_sample_support":true,"all_exclusions_reported":true},
14
+ "regimes": ["source_only","target_64","target_256","source_plus_target_64"],
15
+ "estimation_models": ["support_mixture","joint_jeffreys","naive_bayes"],
16
+ "representations": ["query_joint","posterior_clusters","posterior_bins","future_marginal","raw_information","raw_entropy","hash_code","identity"],
17
+ "planner": {"reuse":"unionmind.decision_sufficiency.core unchanged","utility_correct":1,"utility_wrong":-2,"utility_abstain":0,"query_budgets":[0,1,2],"query_costs":[0.0,0.01,0.05],"cost_interpretation":"modeled feature-read utility cost; not physical sensor cost, transport bytes or engineer time"},
18
+ "primary": {"regime":"target_64","model":"support_mixture","representation":"query_joint","aggregation":"mean over query budgets 1,2, all three costs, three sample seeds, equal tasks within target environment, equal target environments within dataset, equal datasets","compact_controls":["posterior_clusters","posterior_bins","future_marginal","raw_information","raw_entropy","hash_code"],"effect_minimum":0.01,"confidence":"95% simultaneous Bonferroni intervals across six compact controls, paired target-environment block bootstrap within each of the two fixed datasets; tasks, adaptation seeds, costs and budgets are NOT independent replications","bootstrap_draws":10000,"bootstrap_seed":20131,"gate":"gain >= .01 with positive simultaneous lower bound against every compact control, and positive point gain over every compact control in each dataset separately; at least two eligible target environments per dataset"},
19
+ "full_reference_gate": {"comparison":"all three identity models on the same examples","noninferiority_margin":0.01,"confidence":"simultaneous intervals with family of nine comparisons including compact controls"},
20
+ "cost_gate": {"minimum_complete_time_reduction":0.2,"requests":[1,32,1024],"repeats":11,"order":"alternating forward/reverse; common CUDA/library warmup separately charged to all","components":["download","decompression and raw feature extraction","unlabeled source median and event encoding","adaptation sample materialization","counting","model estimation","representation construction","probability pushforward","exact planning","policy replay"],"gate":"20% median complete savings against the fastest full statistical alternative at every request count; scientific gain alone cannot promote a slower algorithm"},
21
+ "fidelity": {"representation":"exact within-fitted-model compression loss and conditional total variation; no claim of true-law sufficiency","estimation":"held-record posterior log loss, Brier, and joint predictive log loss; test-empirical hindsight oracle used only for descriptive decomposition, never deployable inference or a population certificate"},
22
+ "integrity": {"lock_before":"download and opening any numerical source records","retain_all_previous_files":true,"no_threshold_changes_after_opening":true,"all_eligible_tasks_reported":true,"no_test_selected_recipe":true,"sequential_replay":"withheld bits accessed only after the corresponding query action; compare independent event replay to aggregated evaluator","no_human_dependency":true},
23
+ "falsification": ["A failed primary gate rejects external compact-representation improvement for this registered task.","Improvement only at 256 labels is secondary and cannot rescue the 64-label primary claim.","No end-to-end savings rejects computational superiority even if scientific representation benefit is positive.","Two laboratory datasets do not establish production reliability or general transfer across operational environments.","The full reference is the uncompressed eight-binary-feature estimated model, not all raw sensor data."]
24
+ }
protocols/factorial_transfer_protocol_01.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "factorial_transfer_01",
3
+ "objective": "Test a substantial full-hypothesis, scarce-label diagnostic capability: compose previously unseen combinations of equipment conditions and gas levels, estimate their numerical observation mechanisms, and use verified probability and decision operators. Preserve the entire preceding research program.",
4
+ "motivation": "Opened pairwise audits found raw numerical evidence useful, but 8/36-parameter per-task metric learning added no material advantage. The new challenge removes the oracle two-label restriction and uses 64 total fitting examples for the complete ontology, rather than 64 per label pair.",
5
+ "backbone": "This experiment explicitly uses statistical and <=20000-parameter auxiliary world models, without a language backbone or collaborative execution. Original frozen Qwen and cognitive checkpoints remain protected.",
6
+ "device": "cuda:1",
7
+ "analysis_device": "cuda:2",
8
+ "statistical_dtype": "float64",
9
+ "neural_training_dtype": "float32",
10
+ "sources": {
11
+ "hydraulic": {
12
+ "uci": 447,
13
+ "doi": "10.24432/C5CW21",
14
+ "authors": [
15
+ "Nikolai Helwig",
16
+ "Eliseo Pignanelli",
17
+ "Andreas Schuetze"
18
+ ],
19
+ "url": "https://archive.ics.uci.edu/static/public/447/condition+monitoring+of+hydraulic+systems.zip",
20
+ "system": "hydraulic_rig",
21
+ "license": "CC BY 4.0",
22
+ "cycle_seconds": 60,
23
+ "physical_sensors": {
24
+ "PS1": [
25
+ "bar",
26
+ 100
27
+ ],
28
+ "PS2": [
29
+ "bar",
30
+ 100
31
+ ],
32
+ "PS3": [
33
+ "bar",
34
+ 100
35
+ ],
36
+ "PS4": [
37
+ "bar",
38
+ 100
39
+ ],
40
+ "PS5": [
41
+ "bar",
42
+ 100
43
+ ],
44
+ "PS6": [
45
+ "bar",
46
+ 100
47
+ ],
48
+ "EPS1": [
49
+ "W",
50
+ 100
51
+ ],
52
+ "FS1": [
53
+ "l/min",
54
+ 10
55
+ ],
56
+ "FS2": [
57
+ "l/min",
58
+ 10
59
+ ],
60
+ "TS1": [
61
+ "degC",
62
+ 1
63
+ ],
64
+ "TS2": [
65
+ "degC",
66
+ 1
67
+ ],
68
+ "TS3": [
69
+ "degC",
70
+ 1
71
+ ],
72
+ "TS4": [
73
+ "degC",
74
+ 1
75
+ ],
76
+ "VS1": [
77
+ "mm/s",
78
+ 1
79
+ ]
80
+ },
81
+ "excluded_virtual_sensors": [
82
+ "CE",
83
+ "CP",
84
+ "SE"
85
+ ],
86
+ "factor_names": [
87
+ "cooler_condition_percent",
88
+ "valve_condition_percent",
89
+ "pump_leakage_grade",
90
+ "accumulator_pressure_bar"
91
+ ],
92
+ "levels": [
93
+ [
94
+ 3,
95
+ 20,
96
+ 100
97
+ ],
98
+ [
99
+ 73,
100
+ 80,
101
+ 90,
102
+ 100
103
+ ],
104
+ [
105
+ 0,
106
+ 1,
107
+ 2
108
+ ],
109
+ [
110
+ 90,
111
+ 100,
112
+ 115,
113
+ 130
114
+ ]
115
+ ],
116
+ "features": "Mean of each physical channel in all four 15-second cycle quarters, 56 values. Retain all physical channels; no outcome-based feature selection. Stable/transient source flags are reported separately and never exclude a cycle.",
117
+ "record_group": "original five-cycle blocks",
118
+ "opened_archive": "artifacts/external_decision_sufficiency_01/sources/hydraulic.zip"
119
+ },
120
+ "gas": {
121
+ "uci": 322,
122
+ "doi": "10.24432/C5WP4C",
123
+ "authors": [
124
+ "Jordi Fonollosa"
125
+ ],
126
+ "url": "https://archive.ics.uci.edu/static/public/322/gas+sensor+array+under+dynamic+gas+mixtures.zip",
127
+ "system": "gas_array_rig",
128
+ "license": "CC BY 4.0 repository metadata; source description additionally restricts use to research, which is respected",
129
+ "files": {
130
+ "gas_co": {
131
+ "file": "ethylene_CO.txt",
132
+ "primary_max_ppm": 600
133
+ },
134
+ "gas_methane": {
135
+ "file": "ethylene_methane.txt",
136
+ "primary_max_ppm": 300
137
+ }
138
+ },
139
+ "factor_names": [
140
+ "primary_gas_level",
141
+ "ethylene_level"
142
+ ],
143
+ "levels": [
144
+ [
145
+ 0,
146
+ 1,
147
+ 2,
148
+ 3
149
+ ],
150
+ [
151
+ 0,
152
+ 1,
153
+ 2,
154
+ 3
155
+ ]
156
+ ],
157
+ "physical_units": "Timestamp seconds; setpoints ppm; 16 recorded sensor outputs. The author writes the resistance conversion literally as '40.000/S_i'. This experiment retains source readings and does not interpret or perform an unverified resistance-unit conversion.",
158
+ "binning": "Primary: zero, (0,max/4], (max/4,max/2], (max/2,max]; ethylene: zero, (0,5], (5,10], (10,20]. Fixed before data access.",
159
+ "episode": "Maximal contiguous constant setpoint pair; require duration >=75 seconds and strictly increasing timestamps. One episode is one observation, never millions of independent 100-Hz rows.",
160
+ "features": "For every sensor use mean in last 20 seconds and last-20-minus-first-20-second mean, 32 values. No sensor subset. Retain every short-episode exclusion with raw row/time provenance.",
161
+ "record_group": "original exposure episode; two mixtures share a physical system and are not counted as two independent systems"
162
+ }
163
+ },
164
+ "hypothesis_spaces": {
165
+ "hydraulic": 144,
166
+ "gas_co": 16,
167
+ "gas_methane": 16
168
+ },
169
+ "source_access": "Hydraulic archive and labels were opened by external_decision_sufficiency_01. Its new full-ontology development audit is explicitly exploratory. The UCI322 gas-mixture archive, numeric values and labels have not been accessed before this lock. The two fresh experiments share one physical sensor platform: they are not two independently sampled systems.",
170
+ "splits": "Before fitting: sort the entire supplied Cartesian hypothesis ontology by SHA256('20711:'+dataset+':'+comma-separated integer factor indices), holding the first max(4,ceil(.2*ontology_size)) bindings. No sensor value or observed label count chooses bindings. Among other bindings, original record-group hash SHA256('20721:'+dataset+':'+group_index) mod 5==0 is seen-binding held. Source excludes held rows and groups adjacent to any held group. Do not change split seeds after access.",
171
+ "eligibility": "Primary requires >=64 source observations, >=32 unseen-binding held observations in >=3 distinct bindings, and >=16 seen-binding held observations. Secondary 128-label result requires >=128 source observations and is otherwise explicitly unavailable. Keep every series and failure; no replacement or threshold amendment.",
172
+ "sample_budgets": [
173
+ 64,
174
+ 128
175
+ ],
176
+ "sample_seeds": [
177
+ 20751,
178
+ 20752,
179
+ 20753
180
+ ],
181
+ "sampling": "Same random source prefix for every model and both budgets, without replacement, generator seed=sample_seed + series_index*101. No class balancing, label-aware prefix selection or target tuning. Report factor levels absent from each actual fitting prefix and supported/unseen-level strata without dropping cases.",
182
+ "normalization": "Feature mean and standard deviation from the actual 64/128 fitting records only. No full-source label-conditional quantiles. Standard deviation floor 1e-6. All models receive identical normalized numerical evidence.",
183
+ "models": [
184
+ "additive_full",
185
+ "additive_diag",
186
+ "factorized_neural",
187
+ "joint_neural",
188
+ "factorized_nb",
189
+ "factorized_lda",
190
+ "factorized_kernel",
191
+ "factorized_ridge",
192
+ "joint_gaussian",
193
+ "factorized_logistic",
194
+ "factorized_classifier"
195
+ ],
196
+ "candidates": [
197
+ "additive_full",
198
+ "factorized_neural"
199
+ ],
200
+ "world_models": "Additive: all factor one-hots plus intercept, ridge 8, intercept ridge .01, analytic leave-one-out residuals. Factorized neural: full factor one-hots -> SiLU hidden -> observation means. Joint neural: one-hot complete binding -> matched SiLU hidden -> observation means, unknown bindings use uninformative source mean. Hidden widths give <=20000 counted trainable parameters and near-matched factorized/nonfactorized budgets. Both neural decoders fit 400 AdamW steps, lr .003, weight_decay .01; four source folds estimate held residuals plus one full fit, all charged. Source residual covariance is .5 full + .5 diagonal with .05 identity floor; diagonal ablation uses only the diagonal. Priors uniform over the supplied full hypothesis ontology, including unseen bindings.",
201
+ "strong_controls": "Independent-factor Gaussian NB and shared-covariance Gaussian, independent-factor whitened kernel with bandwidth .7 and .5 Dirichlet smoothing, independent-factor ridge classifier with ridge 8, joint-binding Gaussian with shrink-to-source fallback. Use all observed coordinates and the same fitting labels. All factorized controls can predict unseen bindings; do not make the comparison only against memorization. Also include class-balanced multinomial logistic regression and a matched <=20000-parameter SiLU direct classifier, both predicting independent factor probabilities, with the same 400-step optimizer and four out-of-fitting-fold source predictions plus full fit. Their four-point source temperature grid (.5,1,2,4) is chosen only on source cross-fit log loss; every fit is charged. All source- and target-state priors remain uniform.",
202
+ "uncertainty_controls": "No target calibration. Neural observation uncertainty comes from four out-of-fitting-fold source predictions; additive uncertainty uses analytic LOO residuals. Report posterior and conditional observation proper scores, confidence/coverage and source-support gaps. No operational calibration claim without its own validation.",
203
+ "primary": "At 64 total fitting labels, joint Top-1 and Top-3 on all unseen-binding held observations in the two fresh gas mixtures, equal weight per mixture. The previously opened hydraulic rig is an explicit development result and contributes no weight to fresh-confirmation estimates. Seen-binding, unsupported factor levels, all series and both label budgets are mandatory. Top-3 uses uniform tie credit with tolerance1e-12; deterministic results retained.",
204
+ "uncertainty": {
205
+ "draws": 20000,
206
+ "seed": 20791,
207
+ "family": 40,
208
+ "unit": "Shared original held recording groups, same weights across models/seeds. Gas experiments have equal weight. Conditional on one fixed sensor platform and its two experiments, not a population-of-systems interval. Hydraulic estimates remain exploratory.",
209
+ "contrasts": "two frozen candidates versus their ten alternatives at both declared label budgets; unavailable secondary cells retained and no comparison family shrunk"
210
+ },
211
+ "success": {
212
+ "substantial_diagnostic_gain": "At least .10 fresh unseen-binding joint Top-3 gain over every one of the seven strong conventional/direct-classifier controls, positive simultaneous lower bounds, positive point gain in each fresh gas experiment. All three series must meet eligibility for the complete planned study; the opened hydraulic result cannot establish fresh multi-system generalization.",
213
+ "strongest_available": "Positive simultaneous lower bound against every eligible alternative, including matched structured and neural controls. Do not call a method the overall winner if this fails.",
214
+ "learning_specific": "The factorized neural candidate must beat the additive reference with positive simultaneous interval before claiming neural learning adds value.",
215
+ "promotion": "Substantial diagnostic gain plus strongest-available gate, complete compute ledger and intact cohort. A favorable secondary does not erase a failed primary."
216
+ },
217
+ "secondary_acquisition": {
218
+ "world_models": [
219
+ "additive_full",
220
+ "factorized_neural"
221
+ ],
222
+ "cohort": "At most 32 unseen-binding held records per series, ordered by SHA256('20781:'+raw_record_id), frozen before scores; every primary record remains in the diagnostic evaluation.",
223
+ "initial_observed_coordinates": [],
224
+ "budgets": [
225
+ 2,
226
+ 4,
227
+ 8
228
+ ],
229
+ "per_scalar_recorded_query_cost": 0.01,
230
+ "policies": [
231
+ "value_forced",
232
+ "value_stop",
233
+ "variance_forced",
234
+ "variance_stop",
235
+ "static_forced",
236
+ "random_forced"
237
+ ],
238
+ "value": "Exact one-step expected joint Top-3 coverage using shared-variance Gaussian posterior-rank intervals and normal CDFs. No access to an actual withheld value before paying its query. Conditional variance is the strongest cheap adaptive control; variance_stop evaluates only its selected query's decision value. No claim of globally optimal multistep control.",
239
+ "scale_gate": "Either >=.10 Top-3 gain at equal query cap over every cheap control, or >=50% fewer recorded queries with simultaneous noninferiority loss <=.02, plus full computation. This cannot rescue a failed diagnostic or learning-specific gate.",
240
+ "uncertainty": "20000 paired recording-group bootstrap draws, family120, seed20811, two fresh gas experiments equally weighted. Report every policy and both opened/fresh provenance."
241
+ },
242
+ "compute": "CUDA float64 statistics, probability operations, exact interval decisions and bootstrap; float32 auxiliary neural fitting. CPU archive text parsing, provenance, hashing and report aggregation only. Count data preparation, normalization, every full/cross-fit adaptation, full-hypothesis mean materialization, covariance construction, predictions, query decisions, peak GPU memory, trainable/checkpoint/derived-state bytes. No CPU offload, DeepSpeed, Qwen executions or model collaboration.",
243
+ "stopping": "Freeze all code, ontology, features, priors, controls and gates before new gas download. Hydraulic development has been opened and all results remain exploratory. No architecture, threshold, split or favorable subset amendment after new confirmation access. Preserve all failures, previous checkpoints and the evidence engine. No manual review dependency or paper/deployment objective."
244
+ }
provenance/controlled/compute_ledger.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "frozen_code_changed": false,
3
+ "protocol_changed": false,
4
+ "promotion_gates_changed": false,
5
+ "device": "cuda:2",
6
+ "gpu": "NVIDIA H200",
7
+ "model_and_policy_math": "CUDA float64; no CPU offload",
8
+ "backbone_executions": 0,
9
+ "active_backbone_copies": 0,
10
+ "neural_parameters_added": 0,
11
+ "controller_executions": 0,
12
+ "controller_refits": 0,
13
+ "maximum_peak_allocated_bytes": 134986752,
14
+ "maximum_peak_reserved_bytes": 155189248,
15
+ "confirmation_research_seconds": 40.3597513709974,
16
+ "confirmation_scored_records": 1022976,
17
+ "cuda_startup_seconds": 18.617308109998703,
18
+ "cuda_warmups": {
19
+ "support_mixture:query_joint": 0.2577872660185676,
20
+ "joint_jeffreys:identity": 0.0032121709955390543,
21
+ "naive_bayes:identity": 0.003511357994284481,
22
+ "support_mixture:identity": 0.003404008981306106
23
+ },
24
+ "benchmark_records": 176,
25
+ "fitted_statistical_model": {
26
+ "raw_probability_cells": 512,
27
+ "raw_multinomial_degrees_of_freedom": 511,
28
+ "naive_bayes_degrees_of_freedom": 17,
29
+ "support_mixture_weight": "N/(N+512), not fitted",
30
+ "clustering_centers_query_joint": 64,
31
+ "clustering_centers_posterior": 16,
32
+ "centers_retained_at_deployment": false,
33
+ "iterations": 32
34
+ },
35
+ "per_world_state_bytes": {
36
+ "encoded_joint": 512,
37
+ "encoded_lookup": 64,
38
+ "encoded_planner_choices": 5184,
39
+ "encoded_planner_values": 5184,
40
+ "planner_expected_summary": 72,
41
+ "encoded_total_with_planner": 11016,
42
+ "uncompressed_joint": 4096,
43
+ "uncompressed_encoder_lookup": 0,
44
+ "uncompressed_planner_choices": 41472,
45
+ "uncompressed_planner_values": 41472,
46
+ "uncompressed_total_with_planner": 87112,
47
+ "raw_count_table": 4096,
48
+ "original_full_fit_model": 4096,
49
+ "primary_training_records": 8192,
50
+ "measured_primary_live_buffers_including_fit_and_records": 27400,
51
+ "measured_uncompressed_live_buffers_including_fit_and_records": 99400,
52
+ "note": "Tensor payloads; Python object/container and CUDA context overhead excluded from these payload counts and represented by measured allocation/startup separately"
53
+ },
54
+ "accounting": {
55
+ "truth_generation": "Research/evaluation only; never supplied as policy model",
56
+ "statistical_fit_and_encoding": "Charged in construction; all label budgets and adaptation stages in timings.json",
57
+ "probability_materialization": "Charged separately in every stage; dense fit is not claimed free",
58
+ "exact_planning": "Charged, with all hypothetical branches; full true-law oracles are evaluator-only",
59
+ "warmup": "Charged; first alternative includes common library initialization, ordering limitation disclosed",
60
+ "interpretation": "Local batched workload, not production latency, engineering investigation time or evidence-backend throughput"
61
+ },
62
+ "research_stage_timing_source": "confirmation_evaluation/timings.json",
63
+ "all_complete_benchmark_records": "confirmation_benchmark.json"
64
+ }
provenance/controlled/confirmation_benchmark.json ADDED
The diff for this file is too large to render. See raw diff
 
provenance/controlled/preconfirmation_lock.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "unionmind/decision_sufficiency/__init__.py": "f87b50f4b7c6290b319f63d846f1bf8bfc69249818fd2ed8108356b397f43e4d",
4
+ "unionmind/decision_sufficiency/analysis.py": "eeb72ce872762b4af5f19c3d9d39013c05c407f32cb007733f4005fd107377ef",
5
+ "unionmind/decision_sufficiency/benchmark.py": "de9d93e521ee59882f23015fdc8e3912c8b765a412952ba4d16eb5a812722ed9",
6
+ "unionmind/decision_sufficiency/core.py": "fa2f7405a8d0dc6ec96ebae1bb2a2b810d03ef1fd2877f5ebef7df1722992cee",
7
+ "unionmind/decision_sufficiency/study.py": "dcb7d8303fe11e9e1c27c1e8b9a141f4571f49f1c1611a1b3572537e4ce7ec19",
8
+ "unionmind/decision_sufficiency/worlds.py": "b2dd3a0db509161982f8159caeba35f8af5212aef0e3685ac56b702361b15d5e",
9
+ "configs/decision_sufficiency_protocol_01.json": "5a334fe26903403ff48eb71e7b0eb548d139af50c97b430252db0cf71b2c1f59",
10
+ "tests/test_decision_sufficiency.py": "74a02c10008eb360da092b79644960695385dd8b7222e7ef5a1cca7ff9739596",
11
+ "artifacts/decision_sufficiency_01/interface_tests_final.log": "59bece864998a3b86725b1476a948c37f0a02384b05efc32320e749a3b7a6879",
12
+ "artifacts/decision_sufficiency_01/development/worlds.pt": "fb6404a2cf08d1d4e279ca8dc223793933571961c8c3d02224879ccea68f6e69",
13
+ "artifacts/decision_sufficiency_01/development/manifest.json": "663ede689fd468ac008860dfad75c7414edd02b55c3a3c4df7e627bb07719175",
14
+ "artifacts/decision_sufficiency_01/development_evaluation/summary.json": "c3dde81b65b3fd28a25ad89d0b0376eb2b9a4cfbebaaa1cdcfe7ed0c73e3d9a3"
15
+ },
16
+ "confirmation_generated": false,
17
+ "protocol_sha256": "5a334fe26903403ff48eb71e7b0eb548d139af50c97b430252db0cf71b2c1f59",
18
+ "new_controller_training": false,
19
+ "independent_external_confirmation": false,
20
+ "thresholds_and_all_comparisons_fixed": true,
21
+ "test_log": "artifacts/decision_sufficiency_01/interface_tests_final.log"
22
+ }
provenance/external/diagnostic_review.json ADDED
The diff for this file is too large to render. See raw diff
 
provenance/external/preconfirmation_lock.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "unionmind/external_sufficiency/study.py": "7053ad47d088807ecf30af59fccb672dca706eca0d498ecc90a7131d677222c6",
4
+ "unionmind/external_sufficiency/__init__.py": "017b971fb34357fd8b80dfb002e248fafacefda6dbdde8b976b66f350471865f",
5
+ "unionmind/external_sufficiency/data.py": "20275fb0a49a3835c286076ffad5a59312081ef7d1993b9706b873de6d2c3976",
6
+ "unionmind/external_sufficiency/benchmark.py": "a7999a53c7a6ca187912af7b47cfaa5128df3c47610a695e0163131ac05ca5cb",
7
+ "unionmind/external_sufficiency/runtime.py": "fa19ce1b7fba9e6953e9e4081d4b6001ed55c33edc48a05960d5a13ae6aad61d",
8
+ "unionmind/external_sufficiency/analysis.py": "94040f14991c1784122d19f62ececc3f1533641b7b7738966c058b2749033a98",
9
+ "configs/external_sufficiency_protocol_01.json": "e0229ef8a50270b7a590c6ed79a850ec8f902d17f2c8213111bf432669ef9d04",
10
+ "tests/test_external_sufficiency.py": "9bd0aa1b4a546896fb8114561494138a35b79ce85aa921e89a4cd2d6bc68a126",
11
+ "unionmind/decision_sufficiency/core.py": "fa2f7405a8d0dc6ec96ebae1bb2a2b810d03ef1fd2877f5ebef7df1722992cee"
12
+ },
13
+ "locked_before_any_source_download": true,
14
+ "registered_threshold_changes_permitted": false,
15
+ "archived_results_protected": 92804,
16
+ "created_unix": 1791683183.0260608
17
+ }
provenance/external/sources/manifest.json ADDED
@@ -0,0 +1,213 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "archives": [
3
+ {
4
+ "name": "hydraulic",
5
+ "url": "https://archive.ics.uci.edu/static/public/447/condition+monitoring+of+hydraulic+systems.zip",
6
+ "doi": "10.24432/C5CW21",
7
+ "license": "UCI CC BY 4.0",
8
+ "sha256": "24128aad2ee45eea7e6b63ebbd9992cdf25d0483a2cebefbfc13bc69079af1f2",
9
+ "bytes": 76601704,
10
+ "download_seconds": 8.07446516197524,
11
+ "inventory": [
12
+ {
13
+ "name": "CE.txt",
14
+ "uncompressed_bytes": 911516,
15
+ "compressed_bytes": 220629,
16
+ "crc32": 2052641160
17
+ },
18
+ {
19
+ "name": "CP.txt",
20
+ "uncompressed_bytes": 779343,
21
+ "compressed_bytes": 139340,
22
+ "crc32": 2525430905
23
+ },
24
+ {
25
+ "name": "description.txt",
26
+ "uncompressed_bytes": 4044,
27
+ "compressed_bytes": 1752,
28
+ "crc32": 649927169
29
+ },
30
+ {
31
+ "name": "EPS1.txt",
32
+ "uncompressed_bytes": 87411594,
33
+ "compressed_bytes": 5963787,
34
+ "crc32": 2113107188
35
+ },
36
+ {
37
+ "name": "FS1.txt",
38
+ "uncompressed_bytes": 7581926,
39
+ "compressed_bytes": 1822302,
40
+ "crc32": 1100718821
41
+ },
42
+ {
43
+ "name": "FS2.txt",
44
+ "uncompressed_bytes": 8271923,
45
+ "compressed_bytes": 1353203,
46
+ "crc32": 3778132480
47
+ },
48
+ {
49
+ "name": "profile.txt",
50
+ "uncompressed_bytes": 31196,
51
+ "compressed_bytes": 562,
52
+ "crc32": 4140950149
53
+ },
54
+ {
55
+ "name": "PS1.txt",
56
+ "uncompressed_bytes": 91391892,
57
+ "compressed_bytes": 13958273,
58
+ "crc32": 224205924
59
+ },
60
+ {
61
+ "name": "PS2.txt",
62
+ "uncompressed_bytes": 82373611,
63
+ "compressed_bytes": 13268963,
64
+ "crc32": 3777446615
65
+ },
66
+ {
67
+ "name": "PS3.txt",
68
+ "uncompressed_bytes": 69916801,
69
+ "compressed_bytes": 10282091,
70
+ "crc32": 1134355471
71
+ },
72
+ {
73
+ "name": "PS4.txt",
74
+ "uncompressed_bytes": 46698180,
75
+ "compressed_bytes": 5031431,
76
+ "crc32": 3876646838
77
+ },
78
+ {
79
+ "name": "PS5.txt",
80
+ "uncompressed_bytes": 77776429,
81
+ "compressed_bytes": 11765046,
82
+ "crc32": 769051529
83
+ },
84
+ {
85
+ "name": "PS6.txt",
86
+ "uncompressed_bytes": 77875407,
87
+ "compressed_bytes": 11962951,
88
+ "crc32": 1795330865
89
+ },
90
+ {
91
+ "name": "SE.txt",
92
+ "uncompressed_bytes": 824838,
93
+ "compressed_bytes": 164145,
94
+ "crc32": 56425202
95
+ },
96
+ {
97
+ "name": "TS1.txt",
98
+ "uncompressed_bytes": 904586,
99
+ "compressed_bytes": 138072,
100
+ "crc32": 2966138643
101
+ },
102
+ {
103
+ "name": "TS2.txt",
104
+ "uncompressed_bytes": 910594,
105
+ "compressed_bytes": 142687,
106
+ "crc32": 3205193556
107
+ },
108
+ {
109
+ "name": "TS3.txt",
110
+ "uncompressed_bytes": 908805,
111
+ "compressed_bytes": 121394,
112
+ "crc32": 2182737718
113
+ },
114
+ {
115
+ "name": "TS4.txt",
116
+ "uncompressed_bytes": 910127,
117
+ "compressed_bytes": 117953,
118
+ "crc32": 273780249
119
+ },
120
+ {
121
+ "name": "VS1.txt",
122
+ "uncompressed_bytes": 778499,
123
+ "compressed_bytes": 142586,
124
+ "crc32": 1116162816
125
+ },
126
+ {
127
+ "name": "documentation.txt",
128
+ "uncompressed_bytes": 4719,
129
+ "compressed_bytes": 1991,
130
+ "crc32": 37068681
131
+ }
132
+ ]
133
+ },
134
+ {
135
+ "name": "gas",
136
+ "url": "https://archive.ics.uci.edu/static/public/224/gas+sensor+array+drift+dataset.zip",
137
+ "doi": "10.24432/C5RP6W",
138
+ "license": "UCI CC BY 4.0; legacy description additionally restricts use to research, which this study respects",
139
+ "sha256": "91e8f466f202e7a093d657673ce47311c3e90416f7df3057966058961c351fe4",
140
+ "bytes": 9989262,
141
+ "download_seconds": 3.244406032987172,
142
+ "inventory": [
143
+ {
144
+ "name": "Dataset/",
145
+ "uncompressed_bytes": 0,
146
+ "compressed_bytes": 0,
147
+ "crc32": 0
148
+ },
149
+ {
150
+ "name": "Dataset/batch1.dat",
151
+ "uncompressed_bytes": 752726,
152
+ "compressed_bytes": 332626,
153
+ "crc32": 3109529229
154
+ },
155
+ {
156
+ "name": "Dataset/batch10.dat",
157
+ "uncompressed_bytes": 6045679,
158
+ "compressed_bytes": 2661905,
159
+ "crc32": 10376135
160
+ },
161
+ {
162
+ "name": "Dataset/batch2.dat",
163
+ "uncompressed_bytes": 2119601,
164
+ "compressed_bytes": 873108,
165
+ "crc32": 3994276213
166
+ },
167
+ {
168
+ "name": "Dataset/batch3.dat",
169
+ "uncompressed_bytes": 2706554,
170
+ "compressed_bytes": 1077602,
171
+ "crc32": 1496641385
172
+ },
173
+ {
174
+ "name": "Dataset/batch4.dat",
175
+ "uncompressed_bytes": 276473,
176
+ "compressed_bytes": 116866,
177
+ "crc32": 1911836252
178
+ },
179
+ {
180
+ "name": "Dataset/batch5.dat",
181
+ "uncompressed_bytes": 339035,
182
+ "compressed_bytes": 142394,
183
+ "crc32": 694160393
184
+ },
185
+ {
186
+ "name": "Dataset/batch6.dat",
187
+ "uncompressed_bytes": 3902170,
188
+ "compressed_bytes": 1570765,
189
+ "crc32": 3113624330
190
+ },
191
+ {
192
+ "name": "Dataset/batch7.dat",
193
+ "uncompressed_bytes": 6117334,
194
+ "compressed_bytes": 2695549,
195
+ "crc32": 3650852639
196
+ },
197
+ {
198
+ "name": "Dataset/batch8.dat",
199
+ "uncompressed_bytes": 491904,
200
+ "compressed_bytes": 203212,
201
+ "crc32": 153501895
202
+ },
203
+ {
204
+ "name": "Dataset/batch9.dat",
205
+ "uncompressed_bytes": 785870,
206
+ "compressed_bytes": 313603,
207
+ "crc32": 1925951351
208
+ }
209
+ ]
210
+ }
211
+ ],
212
+ "downloaded_after_lock": true
213
+ }
provenance/external/sources/source_definitions.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "hydraulic": {
3
+ "creators": [
4
+ "Nikolai Helwig",
5
+ "Eliseo Pignanelli",
6
+ "Andreas Sch\u00fctze"
7
+ ],
8
+ "dataset_year": 2015,
9
+ "uci_url": "https://archive.ics.uci.edu/dataset/447/condition+monitoring+of+hydraulic+systems",
10
+ "doi": "10.24432/C5CW21",
11
+ "license": "CC BY 4.0",
12
+ "measurements": "60-second cycles of a laboratory hydraulic test rig",
13
+ "selected_units": {
14
+ "PS1": "bar",
15
+ "PS2": "bar",
16
+ "PS3": "bar",
17
+ "PS4": "bar",
18
+ "FS1": "l/min",
19
+ "TS1": "degrees Celsius",
20
+ "EPS1": "W",
21
+ "VS1": "mm/s"
22
+ }
23
+ },
24
+ "gas": {
25
+ "creator": "Alexander Vergara",
26
+ "dataset_year": 2012,
27
+ "uci_url": "https://archive.ics.uci.edu/dataset/224/gas+sensor+array+drift+dataset",
28
+ "doi": "10.24432/C5RP6W",
29
+ "license": "UCI currently displays CC BY 4.0; legacy description says research-only. This noncommercial research respects both.",
30
+ "selected_features": "Second (normalized resistance response) feature of sensors 1 through 8; dimensionless ratios per provider description",
31
+ "temporal_context": "10 provider batches across 36-month drift experiment; batches 1-3 source, batches 6-10 confirmation",
32
+ "recorded_availability_timestamps": false
33
+ },
34
+ "scope": {
35
+ "fresh_to_this_experiment": true,
36
+ "private_unpublished_data": false,
37
+ "provider_defined_sequential_task": false,
38
+ "physical_retrieval_cost_measured": false
39
+ }
40
+ }
provenance/external/task_manifest.json ADDED
@@ -0,0 +1,2608 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "tasks": [
3
+ {
4
+ "index": 0,
5
+ "dataset": "hydraulic",
6
+ "environment": "context1=73",
7
+ "task": "cooler",
8
+ "counts": {
9
+ "source": 1125,
10
+ "adaptation": 150,
11
+ "evaluation": 210
12
+ },
13
+ "evaluation_hypothesis_counts": [
14
+ 60,
15
+ 150
16
+ ],
17
+ "group_counts": [
18
+ 36,
19
+ 15,
20
+ 21
21
+ ],
22
+ "exclusions": [],
23
+ "context_description": "profile column 1; source optimal, target 73",
24
+ "preprocessing_seconds": 0.060412277991417795,
25
+ "thresholds": [
26
+ 158.59539833333335,
27
+ 107.63906383333331,
28
+ 1.7865311666666663,
29
+ 0.0,
30
+ 6.631050000000001,
31
+ 46.34131666666667,
32
+ 2465.404400000001,
33
+ 0.6037166666666667
34
+ ],
35
+ "file": "task_000.pt"
36
+ },
37
+ {
38
+ "index": 1,
39
+ "dataset": "hydraulic",
40
+ "environment": "context1=80",
41
+ "task": "cooler",
42
+ "counts": {
43
+ "source": 1125,
44
+ "adaptation": 230,
45
+ "evaluation": 130
46
+ },
47
+ "evaluation_hypothesis_counts": [
48
+ 20,
49
+ 110
50
+ ],
51
+ "group_counts": [
52
+ 36,
53
+ 23,
54
+ 13
55
+ ],
56
+ "exclusions": [],
57
+ "context_description": "profile column 1; source optimal, target 80",
58
+ "preprocessing_seconds": 0.001348111021798104,
59
+ "thresholds": [
60
+ 158.59539833333335,
61
+ 107.63906383333331,
62
+ 1.7865311666666663,
63
+ 0.0,
64
+ 6.631050000000001,
65
+ 46.34131666666667,
66
+ 2465.404400000001,
67
+ 0.6037166666666667
68
+ ],
69
+ "file": "task_001.pt"
70
+ },
71
+ {
72
+ "index": 2,
73
+ "dataset": "hydraulic",
74
+ "environment": "context1=90",
75
+ "task": "cooler",
76
+ "counts": {
77
+ "source": 1125,
78
+ "adaptation": 250,
79
+ "evaluation": 110
80
+ },
81
+ "evaluation_hypothesis_counts": [
82
+ 20,
83
+ 90
84
+ ],
85
+ "group_counts": [
86
+ 36,
87
+ 25,
88
+ 11
89
+ ],
90
+ "exclusions": [],
91
+ "context_description": "profile column 1; source optimal, target 90",
92
+ "preprocessing_seconds": 0.0011347559920977801,
93
+ "thresholds": [
94
+ 158.59539833333335,
95
+ 107.63906383333331,
96
+ 1.7865311666666663,
97
+ 0.0,
98
+ 6.631050000000001,
99
+ 46.34131666666667,
100
+ 2465.404400000001,
101
+ 0.6037166666666667
102
+ ],
103
+ "file": "task_002.pt"
104
+ },
105
+ {
106
+ "index": 3,
107
+ "dataset": "hydraulic",
108
+ "environment": "context0=3",
109
+ "task": "valve",
110
+ "counts": {
111
+ "source": 741,
112
+ "adaptation": 267,
113
+ "evaluation": 465
114
+ },
115
+ "evaluation_hypothesis_counts": [
116
+ 275,
117
+ 190
118
+ ],
119
+ "group_counts": [
120
+ 48,
121
+ 24,
122
+ 24
123
+ ],
124
+ "exclusions": [],
125
+ "context_description": "profile column 0; source optimal, target 3",
126
+ "preprocessing_seconds": 0.0011218639847356826,
127
+ "thresholds": [
128
+ 160.92307833333334,
129
+ 109.29459166666668,
130
+ 1.987844,
131
+ 10.070700833333333,
132
+ 6.667073333333334,
133
+ 35.888133333333336,
134
+ 2546.545333333334,
135
+ 0.5484666666666667
136
+ ],
137
+ "file": "task_003.pt"
138
+ },
139
+ {
140
+ "index": 4,
141
+ "dataset": "hydraulic",
142
+ "environment": "context0=20",
143
+ "task": "valve",
144
+ "counts": {
145
+ "source": 741,
146
+ "adaptation": 222,
147
+ "evaluation": 510
148
+ },
149
+ "evaluation_hypothesis_counts": [
150
+ 350,
151
+ 160
152
+ ],
153
+ "group_counts": [
154
+ 48,
155
+ 22,
156
+ 26
157
+ ],
158
+ "exclusions": [],
159
+ "context_description": "profile column 0; source optimal, target 20",
160
+ "preprocessing_seconds": 0.0011080769763793796,
161
+ "thresholds": [
162
+ 160.92307833333334,
163
+ 109.29459166666668,
164
+ 1.987844,
165
+ 10.070700833333333,
166
+ 6.667073333333334,
167
+ 35.888133333333336,
168
+ 2546.545333333334,
169
+ 0.5484666666666667
170
+ ],
171
+ "file": "task_004.pt"
172
+ },
173
+ {
174
+ "index": 5,
175
+ "dataset": "hydraulic",
176
+ "environment": "context0=3",
177
+ "task": "pump",
178
+ "counts": {
179
+ "source": 741,
180
+ "adaptation": 267,
181
+ "evaluation": 465
182
+ },
183
+ "evaluation_hypothesis_counts": [
184
+ 302,
185
+ 163
186
+ ],
187
+ "group_counts": [
188
+ 48,
189
+ 24,
190
+ 24
191
+ ],
192
+ "exclusions": [],
193
+ "context_description": "profile column 0; source optimal, target 3",
194
+ "preprocessing_seconds": 0.0010964440007228404,
195
+ "thresholds": [
196
+ 160.92307833333334,
197
+ 109.29459166666668,
198
+ 1.987844,
199
+ 10.070700833333333,
200
+ 6.667073333333334,
201
+ 35.888133333333336,
202
+ 2546.545333333334,
203
+ 0.5484666666666667
204
+ ],
205
+ "file": "task_005.pt"
206
+ },
207
+ {
208
+ "index": 6,
209
+ "dataset": "hydraulic",
210
+ "environment": "context0=20",
211
+ "task": "pump",
212
+ "counts": {
213
+ "source": 741,
214
+ "adaptation": 222,
215
+ "evaluation": 510
216
+ },
217
+ "evaluation_hypothesis_counts": [
218
+ 334,
219
+ 176
220
+ ],
221
+ "group_counts": [
222
+ 48,
223
+ 22,
224
+ 26
225
+ ],
226
+ "exclusions": [],
227
+ "context_description": "profile column 0; source optimal, target 20",
228
+ "preprocessing_seconds": 0.0010927669936791062,
229
+ "thresholds": [
230
+ 160.92307833333334,
231
+ 109.29459166666668,
232
+ 1.987844,
233
+ 10.070700833333333,
234
+ 6.667073333333334,
235
+ 35.888133333333336,
236
+ 2546.545333333334,
237
+ 0.5484666666666667
238
+ ],
239
+ "file": "task_006.pt"
240
+ },
241
+ {
242
+ "index": 7,
243
+ "dataset": "hydraulic",
244
+ "environment": "context0=3",
245
+ "task": "accumulator",
246
+ "counts": {
247
+ "source": 741,
248
+ "adaptation": 267,
249
+ "evaluation": 465
250
+ },
251
+ "evaluation_hypothesis_counts": [
252
+ 262,
253
+ 203
254
+ ],
255
+ "group_counts": [
256
+ 48,
257
+ 24,
258
+ 24
259
+ ],
260
+ "exclusions": [],
261
+ "context_description": "profile column 0; source optimal, target 3",
262
+ "preprocessing_seconds": 0.0010777290153782815,
263
+ "thresholds": [
264
+ 160.92307833333334,
265
+ 109.29459166666668,
266
+ 1.987844,
267
+ 10.070700833333333,
268
+ 6.667073333333334,
269
+ 35.888133333333336,
270
+ 2546.545333333334,
271
+ 0.5484666666666667
272
+ ],
273
+ "file": "task_007.pt"
274
+ },
275
+ {
276
+ "index": 8,
277
+ "dataset": "hydraulic",
278
+ "environment": "context0=20",
279
+ "task": "accumulator",
280
+ "counts": {
281
+ "source": 741,
282
+ "adaptation": 222,
283
+ "evaluation": 510
284
+ },
285
+ "evaluation_hypothesis_counts": [
286
+ 52,
287
+ 458
288
+ ],
289
+ "group_counts": [
290
+ 48,
291
+ 22,
292
+ 26
293
+ ],
294
+ "exclusions": [],
295
+ "context_description": "profile column 0; source optimal, target 20",
296
+ "preprocessing_seconds": 0.0010975780023727566,
297
+ "thresholds": [
298
+ 160.92307833333334,
299
+ 109.29459166666668,
300
+ 1.987844,
301
+ 10.070700833333333,
302
+ 6.667073333333334,
303
+ 35.888133333333336,
304
+ 2546.545333333334,
305
+ 0.5484666666666667
306
+ ],
307
+ "file": "task_008.pt"
308
+ },
309
+ {
310
+ "index": 9,
311
+ "dataset": "gas",
312
+ "environment": "batch6",
313
+ "task": "gas1_vs_2",
314
+ "counts": {
315
+ "source": 1541,
316
+ "adaptation": 608,
317
+ "evaluation": 480
318
+ },
319
+ "evaluation_hypothesis_counts": [
320
+ 277,
321
+ 203
322
+ ],
323
+ "group_counts": [
324
+ 101,
325
+ 39,
326
+ 30
327
+ ],
328
+ "exclusions": [],
329
+ "context_description": "provider temporal batch; source batches 1,2,3",
330
+ "preprocessing_seconds": 0.003434196987655014,
331
+ "thresholds": [
332
+ 3.509948,
333
+ 3.561074,
334
+ 9.307478,
335
+ 9.057845,
336
+ 3.572164,
337
+ 3.519498,
338
+ 10.115748,
339
+ 10.035546
340
+ ],
341
+ "file": "task_009.pt"
342
+ },
343
+ {
344
+ "index": 10,
345
+ "dataset": "gas",
346
+ "environment": "batch7",
347
+ "task": "gas1_vs_2",
348
+ "counts": {
349
+ "source": 1541,
350
+ "adaptation": 683,
351
+ "evaluation": 628
352
+ },
353
+ "evaluation_hypothesis_counts": [
354
+ 306,
355
+ 322
356
+ ],
357
+ "group_counts": [
358
+ 101,
359
+ 63,
360
+ 56
361
+ ],
362
+ "exclusions": [],
363
+ "context_description": "provider temporal batch; source batches 1,2,3",
364
+ "preprocessing_seconds": 0.0053831760014873,
365
+ "thresholds": [
366
+ 3.509948,
367
+ 3.561074,
368
+ 9.307478,
369
+ 9.057845,
370
+ 3.572164,
371
+ 3.519498,
372
+ 10.115748,
373
+ 10.035546
374
+ ],
375
+ "file": "task_010.pt"
376
+ },
377
+ {
378
+ "index": 13,
379
+ "dataset": "gas",
380
+ "environment": "batch10",
381
+ "task": "gas1_vs_2",
382
+ "counts": {
383
+ "source": 1541,
384
+ "adaptation": 620,
385
+ "evaluation": 580
386
+ },
387
+ "evaluation_hypothesis_counts": [
388
+ 268,
389
+ 312
390
+ ],
391
+ "group_counts": [
392
+ 101,
393
+ 59,
394
+ 62
395
+ ],
396
+ "exclusions": [],
397
+ "context_description": "provider temporal batch; source batches 1,2,3",
398
+ "preprocessing_seconds": 0.0032183389994315803,
399
+ "thresholds": [
400
+ 3.509948,
401
+ 3.561074,
402
+ 9.307478,
403
+ 9.057845,
404
+ 3.572164,
405
+ 3.519498,
406
+ 10.115748,
407
+ 10.035546
408
+ ],
409
+ "file": "task_013.pt"
410
+ },
411
+ {
412
+ "index": 14,
413
+ "dataset": "gas",
414
+ "environment": "batch6",
415
+ "task": "gas1_vs_3",
416
+ "counts": {
417
+ "source": 1018,
418
+ "adaptation": 283,
419
+ "evaluation": 341
420
+ },
421
+ "evaluation_hypothesis_counts": [
422
+ 277,
423
+ 64
424
+ ],
425
+ "group_counts": [
426
+ 70,
427
+ 19,
428
+ 22
429
+ ],
430
+ "exclusions": [],
431
+ "context_description": "provider temporal batch; source batches 1,2,3",
432
+ "preprocessing_seconds": 0.0034160409995820373,
433
+ "thresholds": [
434
+ 9.2095,
435
+ 8.154736,
436
+ 5.092842,
437
+ 5.394068,
438
+ 3.628032,
439
+ 3.571421,
440
+ 5.535248,
441
+ 5.678917
442
+ ],
443
+ "file": "task_014.pt"
444
+ },
445
+ {
446
+ "index": 15,
447
+ "dataset": "gas",
448
+ "environment": "batch7",
449
+ "task": "gas1_vs_3",
450
+ "counts": {
451
+ "source": 1018,
452
+ "adaptation": 501,
453
+ "evaluation": 508
454
+ },
455
+ "evaluation_hypothesis_counts": [
456
+ 306,
457
+ 202
458
+ ],
459
+ "group_counts": [
460
+ 70,
461
+ 48,
462
+ 53
463
+ ],
464
+ "exclusions": [],
465
+ "context_description": "provider temporal batch; source batches 1,2,3",
466
+ "preprocessing_seconds": 0.0032475720217917114,
467
+ "thresholds": [
468
+ 9.2095,
469
+ 8.154736,
470
+ 5.092842,
471
+ 5.394068,
472
+ 3.628032,
473
+ 3.571421,
474
+ 5.535248,
475
+ 5.678917
476
+ ],
477
+ "file": "task_015.pt"
478
+ },
479
+ {
480
+ "index": 17,
481
+ "dataset": "gas",
482
+ "environment": "batch9",
483
+ "task": "gas1_vs_3",
484
+ "counts": {
485
+ "source": 1018,
486
+ "adaptation": 71,
487
+ "evaluation": 90
488
+ },
489
+ "evaluation_hypothesis_counts": [
490
+ 42,
491
+ 48
492
+ ],
493
+ "group_counts": [
494
+ 70,
495
+ 6,
496
+ 6
497
+ ],
498
+ "exclusions": [],
499
+ "context_description": "provider temporal batch; source batches 1,2,3",
500
+ "preprocessing_seconds": 0.0030187329975888133,
501
+ "thresholds": [
502
+ 9.2095,
503
+ 8.154736,
504
+ 5.092842,
505
+ 5.394068,
506
+ 3.628032,
507
+ 3.571421,
508
+ 5.535248,
509
+ 5.678917
510
+ ],
511
+ "file": "task_017.pt"
512
+ },
513
+ {
514
+ "index": 18,
515
+ "dataset": "gas",
516
+ "environment": "batch10",
517
+ "task": "gas1_vs_3",
518
+ "counts": {
519
+ "source": 1018,
520
+ "adaptation": 624,
521
+ "evaluation": 576
522
+ },
523
+ "evaluation_hypothesis_counts": [
524
+ 268,
525
+ 308
526
+ ],
527
+ "group_counts": [
528
+ 70,
529
+ 79,
530
+ 74
531
+ ],
532
+ "exclusions": [],
533
+ "context_description": "provider temporal batch; source batches 1,2,3",
534
+ "preprocessing_seconds": 0.0032807049865368754,
535
+ "thresholds": [
536
+ 9.2095,
537
+ 8.154736,
538
+ 5.092842,
539
+ 5.394068,
540
+ 3.628032,
541
+ 3.571421,
542
+ 5.535248,
543
+ 5.678917
544
+ ],
545
+ "file": "task_018.pt"
546
+ },
547
+ {
548
+ "index": 20,
549
+ "dataset": "gas",
550
+ "environment": "batch7",
551
+ "task": "gas1_vs_4",
552
+ "counts": {
553
+ "source": 998,
554
+ "adaptation": 783,
555
+ "evaluation": 610
556
+ },
557
+ "evaluation_hypothesis_counts": [
558
+ 306,
559
+ 304
560
+ ],
561
+ "group_counts": [
562
+ 71,
563
+ 90,
564
+ 71
565
+ ],
566
+ "exclusions": [],
567
+ "context_description": "provider temporal batch; source batches 1,2,3",
568
+ "preprocessing_seconds": 0.002904932975070551,
569
+ "thresholds": [
570
+ 12.177764,
571
+ 10.395643,
572
+ 6.598816,
573
+ 6.345723,
574
+ 3.535424,
575
+ 3.437703,
576
+ 7.377568,
577
+ 7.300357
578
+ ],
579
+ "file": "task_020.pt"
580
+ },
581
+ {
582
+ "index": 23,
583
+ "dataset": "gas",
584
+ "environment": "batch10",
585
+ "task": "gas1_vs_4",
586
+ "counts": {
587
+ "source": 998,
588
+ "adaptation": 648,
589
+ "evaluation": 552
590
+ },
591
+ "evaluation_hypothesis_counts": [
592
+ 268,
593
+ 284
594
+ ],
595
+ "group_counts": [
596
+ 71,
597
+ 82,
598
+ 71
599
+ ],
600
+ "exclusions": [],
601
+ "context_description": "provider temporal batch; source batches 1,2,3",
602
+ "preprocessing_seconds": 0.003192022006260231,
603
+ "thresholds": [
604
+ 12.177764,
605
+ 10.395643,
606
+ 6.598816,
607
+ 6.345723,
608
+ 3.535424,
609
+ 3.437703,
610
+ 7.377568,
611
+ 7.300357
612
+ ],
613
+ "file": "task_023.pt"
614
+ },
615
+ {
616
+ "index": 24,
617
+ "dataset": "gas",
618
+ "environment": "batch6",
619
+ "task": "gas1_vs_5",
620
+ "counts": {
621
+ "source": 1496,
622
+ "adaptation": 639,
623
+ "evaluation": 481
624
+ },
625
+ "evaluation_hypothesis_counts": [
626
+ 277,
627
+ 204
628
+ ],
629
+ "group_counts": [
630
+ 103,
631
+ 41,
632
+ 32
633
+ ],
634
+ "exclusions": [],
635
+ "context_description": "provider temporal batch; source batches 1,2,3",
636
+ "preprocessing_seconds": 0.003304001991637051,
637
+ "thresholds": [
638
+ 6.444352,
639
+ 5.350785,
640
+ 5.635703,
641
+ 5.382907,
642
+ 2.996164,
643
+ 2.94327,
644
+ 6.020208,
645
+ 5.94135
646
+ ],
647
+ "file": "task_024.pt"
648
+ },
649
+ {
650
+ "index": 25,
651
+ "dataset": "gas",
652
+ "environment": "batch7",
653
+ "task": "gas1_vs_5",
654
+ "counts": {
655
+ "source": 1496,
656
+ "adaptation": 726,
657
+ "evaluation": 553
658
+ },
659
+ "evaluation_hypothesis_counts": [
660
+ 306,
661
+ 247
662
+ ],
663
+ "group_counts": [
664
+ 103,
665
+ 83,
666
+ 63
667
+ ],
668
+ "exclusions": [],
669
+ "context_description": "provider temporal batch; source batches 1,2,3",
670
+ "preprocessing_seconds": 0.003856750001432374,
671
+ "thresholds": [
672
+ 6.444352,
673
+ 5.350785,
674
+ 5.635703,
675
+ 5.382907,
676
+ 2.996164,
677
+ 2.94327,
678
+ 6.020208,
679
+ 5.94135
680
+ ],
681
+ "file": "task_025.pt"
682
+ },
683
+ {
684
+ "index": 26,
685
+ "dataset": "gas",
686
+ "environment": "batch8",
687
+ "task": "gas1_vs_5",
688
+ "counts": {
689
+ "source": 1496,
690
+ "adaptation": 67,
691
+ "evaluation": 106
692
+ },
693
+ "evaluation_hypothesis_counts": [
694
+ 11,
695
+ 95
696
+ ],
697
+ "group_counts": [
698
+ 103,
699
+ 7,
700
+ 10
701
+ ],
702
+ "exclusions": [],
703
+ "context_description": "provider temporal batch; source batches 1,2,3",
704
+ "preprocessing_seconds": 0.002878277999116108,
705
+ "thresholds": [
706
+ 6.444352,
707
+ 5.350785,
708
+ 5.635703,
709
+ 5.382907,
710
+ 2.996164,
711
+ 2.94327,
712
+ 6.020208,
713
+ 5.94135
714
+ ],
715
+ "file": "task_026.pt"
716
+ },
717
+ {
718
+ "index": 28,
719
+ "dataset": "gas",
720
+ "environment": "batch10",
721
+ "task": "gas1_vs_5",
722
+ "counts": {
723
+ "source": 1496,
724
+ "adaptation": 632,
725
+ "evaluation": 568
726
+ },
727
+ "evaluation_hypothesis_counts": [
728
+ 268,
729
+ 300
730
+ ],
731
+ "group_counts": [
732
+ 103,
733
+ 57,
734
+ 50
735
+ ],
736
+ "exclusions": [],
737
+ "context_description": "provider temporal batch; source batches 1,2,3",
738
+ "preprocessing_seconds": 0.003230970003642142,
739
+ "thresholds": [
740
+ 6.444352,
741
+ 5.350785,
742
+ 5.635703,
743
+ 5.382907,
744
+ 2.996164,
745
+ 2.94327,
746
+ 6.020208,
747
+ 5.94135
748
+ ],
749
+ "file": "task_028.pt"
750
+ },
751
+ {
752
+ "index": 29,
753
+ "dataset": "gas",
754
+ "environment": "batch6",
755
+ "task": "gas1_vs_6",
756
+ "counts": {
757
+ "source": 698,
758
+ "adaptation": 524,
759
+ "evaluation": 457
760
+ },
761
+ "evaluation_hypothesis_counts": [
762
+ 277,
763
+ 180
764
+ ],
765
+ "group_counts": [
766
+ 50,
767
+ 33,
768
+ 30
769
+ ],
770
+ "exclusions": [],
771
+ "context_description": "provider temporal batch; source batches 1,2,3",
772
+ "preprocessing_seconds": 0.0031620580120943487,
773
+ "thresholds": [
774
+ 18.90034,
775
+ 15.765933,
776
+ 7.043113,
777
+ 6.925369,
778
+ 4.117601,
779
+ 3.988784,
780
+ 8.387417,
781
+ 8.297722
782
+ ],
783
+ "file": "task_029.pt"
784
+ },
785
+ {
786
+ "index": 30,
787
+ "dataset": "gas",
788
+ "environment": "batch7",
789
+ "task": "gas1_vs_6",
790
+ "counts": {
791
+ "source": 698,
792
+ "adaptation": 599,
793
+ "evaluation": 618
794
+ },
795
+ "evaluation_hypothesis_counts": [
796
+ 306,
797
+ 312
798
+ ],
799
+ "group_counts": [
800
+ 50,
801
+ 70,
802
+ 68
803
+ ],
804
+ "exclusions": [],
805
+ "context_description": "provider temporal batch; source batches 1,2,3",
806
+ "preprocessing_seconds": 0.0028372439846862108,
807
+ "thresholds": [
808
+ 18.90034,
809
+ 15.765933,
810
+ 7.043113,
811
+ 6.925369,
812
+ 4.117601,
813
+ 3.988784,
814
+ 8.387417,
815
+ 8.297722
816
+ ],
817
+ "file": "task_030.pt"
818
+ },
819
+ {
820
+ "index": 32,
821
+ "dataset": "gas",
822
+ "environment": "batch9",
823
+ "task": "gas1_vs_6",
824
+ "counts": {
825
+ "source": 698,
826
+ "adaptation": 71,
827
+ "evaluation": 91
828
+ },
829
+ "evaluation_hypothesis_counts": [
830
+ 42,
831
+ 49
832
+ ],
833
+ "group_counts": [
834
+ 50,
835
+ 6,
836
+ 6
837
+ ],
838
+ "exclusions": [],
839
+ "context_description": "provider temporal batch; source batches 1,2,3",
840
+ "preprocessing_seconds": 0.002637180994497612,
841
+ "thresholds": [
842
+ 18.90034,
843
+ 15.765933,
844
+ 7.043113,
845
+ 6.925369,
846
+ 4.117601,
847
+ 3.988784,
848
+ 8.387417,
849
+ 8.297722
850
+ ],
851
+ "file": "task_032.pt"
852
+ },
853
+ {
854
+ "index": 33,
855
+ "dataset": "gas",
856
+ "environment": "batch10",
857
+ "task": "gas1_vs_6",
858
+ "counts": {
859
+ "source": 698,
860
+ "adaptation": 564,
861
+ "evaluation": 636
862
+ },
863
+ "evaluation_hypothesis_counts": [
864
+ 268,
865
+ 368
866
+ ],
867
+ "group_counts": [
868
+ 50,
869
+ 71,
870
+ 81
871
+ ],
872
+ "exclusions": [],
873
+ "context_description": "provider temporal batch; source batches 1,2,3",
874
+ "preprocessing_seconds": 0.0031322890135925263,
875
+ "thresholds": [
876
+ 18.90034,
877
+ 15.765933,
878
+ 7.043113,
879
+ 6.925369,
880
+ 4.117601,
881
+ 3.988784,
882
+ 8.387417,
883
+ 8.297722
884
+ ],
885
+ "file": "task_033.pt"
886
+ },
887
+ {
888
+ "index": 34,
889
+ "dataset": "gas",
890
+ "environment": "batch6",
891
+ "task": "gas2_vs_3",
892
+ "counts": {
893
+ "source": 1321,
894
+ "adaptation": 417,
895
+ "evaluation": 267
896
+ },
897
+ "evaluation_hypothesis_counts": [
898
+ 203,
899
+ 64
900
+ ],
901
+ "group_counts": [
902
+ 89,
903
+ 27,
904
+ 19
905
+ ],
906
+ "exclusions": [],
907
+ "context_description": "provider temporal batch; source batches 1,2,3",
908
+ "preprocessing_seconds": 0.00323677301639691,
909
+ "thresholds": [
910
+ 2.689076,
911
+ 2.808401,
912
+ 8.393048,
913
+ 8.150495,
914
+ 3.049182,
915
+ 3.136738,
916
+ 8.543316,
917
+ 8.654631
918
+ ],
919
+ "file": "task_034.pt"
920
+ },
921
+ {
922
+ "index": 35,
923
+ "dataset": "gas",
924
+ "environment": "batch7",
925
+ "task": "gas2_vs_3",
926
+ "counts": {
927
+ "source": 1321,
928
+ "adaptation": 498,
929
+ "evaluation": 524
930
+ },
931
+ "evaluation_hypothesis_counts": [
932
+ 322,
933
+ 202
934
+ ],
935
+ "group_counts": [
936
+ 89,
937
+ 71,
938
+ 75
939
+ ],
940
+ "exclusions": [],
941
+ "context_description": "provider temporal batch; source batches 1,2,3",
942
+ "preprocessing_seconds": 0.003257372009102255,
943
+ "thresholds": [
944
+ 2.689076,
945
+ 2.808401,
946
+ 8.393048,
947
+ 8.150495,
948
+ 3.049182,
949
+ 3.136738,
950
+ 8.543316,
951
+ 8.654631
952
+ ],
953
+ "file": "task_035.pt"
954
+ },
955
+ {
956
+ "index": 37,
957
+ "dataset": "gas",
958
+ "environment": "batch9",
959
+ "task": "gas2_vs_3",
960
+ "counts": {
961
+ "source": 1321,
962
+ "adaptation": 91,
963
+ "evaluation": 64
964
+ },
965
+ "evaluation_hypothesis_counts": [
966
+ 16,
967
+ 48
968
+ ],
969
+ "group_counts": [
970
+ 89,
971
+ 6,
972
+ 4
973
+ ],
974
+ "exclusions": [],
975
+ "context_description": "provider temporal batch; source batches 1,2,3",
976
+ "preprocessing_seconds": 0.0030166440119501203,
977
+ "thresholds": [
978
+ 2.689076,
979
+ 2.808401,
980
+ 8.393048,
981
+ 8.150495,
982
+ 3.049182,
983
+ 3.136738,
984
+ 8.543316,
985
+ 8.654631
986
+ ],
987
+ "file": "task_037.pt"
988
+ },
989
+ {
990
+ "index": 38,
991
+ "dataset": "gas",
992
+ "environment": "batch10",
993
+ "task": "gas2_vs_3",
994
+ "counts": {
995
+ "source": 1321,
996
+ "adaptation": 580,
997
+ "evaluation": 620
998
+ },
999
+ "evaluation_hypothesis_counts": [
1000
+ 312,
1001
+ 308
1002
+ ],
1003
+ "group_counts": [
1004
+ 89,
1005
+ 73,
1006
+ 79
1007
+ ],
1008
+ "exclusions": [],
1009
+ "context_description": "provider temporal batch; source batches 1,2,3",
1010
+ "preprocessing_seconds": 0.003239960002247244,
1011
+ "thresholds": [
1012
+ 2.689076,
1013
+ 2.808401,
1014
+ 8.393048,
1015
+ 8.150495,
1016
+ 3.049182,
1017
+ 3.136738,
1018
+ 8.543316,
1019
+ 8.654631
1020
+ ],
1021
+ "file": "task_038.pt"
1022
+ },
1023
+ {
1024
+ "index": 40,
1025
+ "dataset": "gas",
1026
+ "environment": "batch7",
1027
+ "task": "gas2_vs_4",
1028
+ "counts": {
1029
+ "source": 1301,
1030
+ "adaptation": 780,
1031
+ "evaluation": 626
1032
+ },
1033
+ "evaluation_hypothesis_counts": [
1034
+ 322,
1035
+ 304
1036
+ ],
1037
+ "group_counts": [
1038
+ 90,
1039
+ 72,
1040
+ 57
1041
+ ],
1042
+ "exclusions": [],
1043
+ "context_description": "provider temporal batch; source batches 1,2,3",
1044
+ "preprocessing_seconds": 0.0029914610204286873,
1045
+ "thresholds": [
1046
+ 2.986544,
1047
+ 3.220806,
1048
+ 8.521594,
1049
+ 8.301294,
1050
+ 3.110722,
1051
+ 3.12992,
1052
+ 9.173403,
1053
+ 9.214573
1054
+ ],
1055
+ "file": "task_040.pt"
1056
+ },
1057
+ {
1058
+ "index": 43,
1059
+ "dataset": "gas",
1060
+ "environment": "batch10",
1061
+ "task": "gas2_vs_4",
1062
+ "counts": {
1063
+ "source": 1301,
1064
+ "adaptation": 604,
1065
+ "evaluation": 596
1066
+ },
1067
+ "evaluation_hypothesis_counts": [
1068
+ 312,
1069
+ 284
1070
+ ],
1071
+ "group_counts": [
1072
+ 90,
1073
+ 76,
1074
+ 77
1075
+ ],
1076
+ "exclusions": [],
1077
+ "context_description": "provider temporal batch; source batches 1,2,3",
1078
+ "preprocessing_seconds": 0.04906752199167386,
1079
+ "thresholds": [
1080
+ 2.986544,
1081
+ 3.220806,
1082
+ 8.521594,
1083
+ 8.301294,
1084
+ 3.110722,
1085
+ 3.12992,
1086
+ 9.173403,
1087
+ 9.214573
1088
+ ],
1089
+ "file": "task_043.pt"
1090
+ },
1091
+ {
1092
+ "index": 44,
1093
+ "dataset": "gas",
1094
+ "environment": "batch6",
1095
+ "task": "gas2_vs_5",
1096
+ "counts": {
1097
+ "source": 1799,
1098
+ "adaptation": 773,
1099
+ "evaluation": 407
1100
+ },
1101
+ "evaluation_hypothesis_counts": [
1102
+ 203,
1103
+ 204
1104
+ ],
1105
+ "group_counts": [
1106
+ 125,
1107
+ 50,
1108
+ 29
1109
+ ],
1110
+ "exclusions": [],
1111
+ "context_description": "provider temporal batch; source batches 1,2,3",
1112
+ "preprocessing_seconds": 0.0033374330087099224,
1113
+ "thresholds": [
1114
+ 3.680179,
1115
+ 3.495189,
1116
+ 6.469984,
1117
+ 6.182178,
1118
+ 2.876749,
1119
+ 2.866318,
1120
+ 6.787725,
1121
+ 6.714887
1122
+ ],
1123
+ "file": "task_044.pt"
1124
+ },
1125
+ {
1126
+ "index": 45,
1127
+ "dataset": "gas",
1128
+ "environment": "batch7",
1129
+ "task": "gas2_vs_5",
1130
+ "counts": {
1131
+ "source": 1799,
1132
+ "adaptation": 723,
1133
+ "evaluation": 569
1134
+ },
1135
+ "evaluation_hypothesis_counts": [
1136
+ 322,
1137
+ 247
1138
+ ],
1139
+ "group_counts": [
1140
+ 125,
1141
+ 89,
1142
+ 71
1143
+ ],
1144
+ "exclusions": [],
1145
+ "context_description": "provider temporal batch; source batches 1,2,3",
1146
+ "preprocessing_seconds": 0.0033002249838318676,
1147
+ "thresholds": [
1148
+ 3.680179,
1149
+ 3.495189,
1150
+ 6.469984,
1151
+ 6.182178,
1152
+ 2.876749,
1153
+ 2.866318,
1154
+ 6.787725,
1155
+ 6.714887
1156
+ ],
1157
+ "file": "task_045.pt"
1158
+ },
1159
+ {
1160
+ "index": 46,
1161
+ "dataset": "gas",
1162
+ "environment": "batch8",
1163
+ "task": "gas2_vs_5",
1164
+ "counts": {
1165
+ "source": 1799,
1166
+ "adaptation": 67,
1167
+ "evaluation": 106
1168
+ },
1169
+ "evaluation_hypothesis_counts": [
1170
+ 11,
1171
+ 95
1172
+ ],
1173
+ "group_counts": [
1174
+ 125,
1175
+ 7,
1176
+ 10
1177
+ ],
1178
+ "exclusions": [],
1179
+ "context_description": "provider temporal batch; source batches 1,2,3",
1180
+ "preprocessing_seconds": 0.0028766260074917227,
1181
+ "thresholds": [
1182
+ 3.680179,
1183
+ 3.495189,
1184
+ 6.469984,
1185
+ 6.182178,
1186
+ 2.876749,
1187
+ 2.866318,
1188
+ 6.787725,
1189
+ 6.714887
1190
+ ],
1191
+ "file": "task_046.pt"
1192
+ },
1193
+ {
1194
+ "index": 48,
1195
+ "dataset": "gas",
1196
+ "environment": "batch10",
1197
+ "task": "gas2_vs_5",
1198
+ "counts": {
1199
+ "source": 1799,
1200
+ "adaptation": 588,
1201
+ "evaluation": 612
1202
+ },
1203
+ "evaluation_hypothesis_counts": [
1204
+ 312,
1205
+ 300
1206
+ ],
1207
+ "group_counts": [
1208
+ 125,
1209
+ 74,
1210
+ 77
1211
+ ],
1212
+ "exclusions": [],
1213
+ "context_description": "provider temporal batch; source batches 1,2,3",
1214
+ "preprocessing_seconds": 0.0031813450041227043,
1215
+ "thresholds": [
1216
+ 3.680179,
1217
+ 3.495189,
1218
+ 6.469984,
1219
+ 6.182178,
1220
+ 2.876749,
1221
+ 2.866318,
1222
+ 6.787725,
1223
+ 6.714887
1224
+ ],
1225
+ "file": "task_048.pt"
1226
+ },
1227
+ {
1228
+ "index": 49,
1229
+ "dataset": "gas",
1230
+ "environment": "batch6",
1231
+ "task": "gas2_vs_6",
1232
+ "counts": {
1233
+ "source": 1001,
1234
+ "adaptation": 658,
1235
+ "evaluation": 383
1236
+ },
1237
+ "evaluation_hypothesis_counts": [
1238
+ 203,
1239
+ 180
1240
+ ],
1241
+ "group_counts": [
1242
+ 71,
1243
+ 43,
1244
+ 27
1245
+ ],
1246
+ "exclusions": [],
1247
+ "context_description": "provider temporal batch; source batches 1,2,3",
1248
+ "preprocessing_seconds": 0.0032896840129978955,
1249
+ "thresholds": [
1250
+ 2.492576,
1251
+ 2.553068,
1252
+ 10.416205,
1253
+ 10.088791,
1254
+ 3.116647,
1255
+ 3.223903,
1256
+ 10.877342,
1257
+ 11.002249
1258
+ ],
1259
+ "file": "task_049.pt"
1260
+ },
1261
+ {
1262
+ "index": 50,
1263
+ "dataset": "gas",
1264
+ "environment": "batch7",
1265
+ "task": "gas2_vs_6",
1266
+ "counts": {
1267
+ "source": 1001,
1268
+ "adaptation": 596,
1269
+ "evaluation": 634
1270
+ },
1271
+ "evaluation_hypothesis_counts": [
1272
+ 322,
1273
+ 312
1274
+ ],
1275
+ "group_counts": [
1276
+ 71,
1277
+ 77,
1278
+ 77
1279
+ ],
1280
+ "exclusions": [],
1281
+ "context_description": "provider temporal batch; source batches 1,2,3",
1282
+ "preprocessing_seconds": 0.0031988440023269504,
1283
+ "thresholds": [
1284
+ 2.492576,
1285
+ 2.553068,
1286
+ 10.416205,
1287
+ 10.088791,
1288
+ 3.116647,
1289
+ 3.223903,
1290
+ 10.877342,
1291
+ 11.002249
1292
+ ],
1293
+ "file": "task_050.pt"
1294
+ },
1295
+ {
1296
+ "index": 52,
1297
+ "dataset": "gas",
1298
+ "environment": "batch9",
1299
+ "task": "gas2_vs_6",
1300
+ "counts": {
1301
+ "source": 1001,
1302
+ "adaptation": 91,
1303
+ "evaluation": 65
1304
+ },
1305
+ "evaluation_hypothesis_counts": [
1306
+ 16,
1307
+ 49
1308
+ ],
1309
+ "group_counts": [
1310
+ 71,
1311
+ 7,
1312
+ 5
1313
+ ],
1314
+ "exclusions": [],
1315
+ "context_description": "provider temporal batch; source batches 1,2,3",
1316
+ "preprocessing_seconds": 0.003260944999055937,
1317
+ "thresholds": [
1318
+ 2.492576,
1319
+ 2.553068,
1320
+ 10.416205,
1321
+ 10.088791,
1322
+ 3.116647,
1323
+ 3.223903,
1324
+ 10.877342,
1325
+ 11.002249
1326
+ ],
1327
+ "file": "task_052.pt"
1328
+ },
1329
+ {
1330
+ "index": 53,
1331
+ "dataset": "gas",
1332
+ "environment": "batch10",
1333
+ "task": "gas2_vs_6",
1334
+ "counts": {
1335
+ "source": 1001,
1336
+ "adaptation": 520,
1337
+ "evaluation": 680
1338
+ },
1339
+ "evaluation_hypothesis_counts": [
1340
+ 312,
1341
+ 368
1342
+ ],
1343
+ "group_counts": [
1344
+ 71,
1345
+ 49,
1346
+ 59
1347
+ ],
1348
+ "exclusions": [],
1349
+ "context_description": "provider temporal batch; source batches 1,2,3",
1350
+ "preprocessing_seconds": 0.003209536982467398,
1351
+ "thresholds": [
1352
+ 2.492576,
1353
+ 2.553068,
1354
+ 10.416205,
1355
+ 10.088791,
1356
+ 3.116647,
1357
+ 3.223903,
1358
+ 10.877342,
1359
+ 11.002249
1360
+ ],
1361
+ "file": "task_053.pt"
1362
+ },
1363
+ {
1364
+ "index": 55,
1365
+ "dataset": "gas",
1366
+ "environment": "batch7",
1367
+ "task": "gas3_vs_4",
1368
+ "counts": {
1369
+ "source": 778,
1370
+ "adaptation": 598,
1371
+ "evaluation": 506
1372
+ },
1373
+ "evaluation_hypothesis_counts": [
1374
+ 202,
1375
+ 304
1376
+ ],
1377
+ "group_counts": [
1378
+ 53,
1379
+ 80,
1380
+ 71
1381
+ ],
1382
+ "exclusions": [],
1383
+ "context_description": "provider temporal batch; source batches 1,2,3",
1384
+ "preprocessing_seconds": 0.003091054008109495,
1385
+ "thresholds": [
1386
+ 6.178135,
1387
+ 5.88016,
1388
+ 4.248013,
1389
+ 4.465726,
1390
+ 2.736321,
1391
+ 2.704917,
1392
+ 4.216443,
1393
+ 4.26598
1394
+ ],
1395
+ "file": "task_055.pt"
1396
+ },
1397
+ {
1398
+ "index": 58,
1399
+ "dataset": "gas",
1400
+ "environment": "batch10",
1401
+ "task": "gas3_vs_4",
1402
+ "counts": {
1403
+ "source": 778,
1404
+ "adaptation": 608,
1405
+ "evaluation": 592
1406
+ },
1407
+ "evaluation_hypothesis_counts": [
1408
+ 308,
1409
+ 284
1410
+ ],
1411
+ "group_counts": [
1412
+ 53,
1413
+ 52,
1414
+ 54
1415
+ ],
1416
+ "exclusions": [],
1417
+ "context_description": "provider temporal batch; source batches 1,2,3",
1418
+ "preprocessing_seconds": 0.003067053999984637,
1419
+ "thresholds": [
1420
+ 6.178135,
1421
+ 5.88016,
1422
+ 4.248013,
1423
+ 4.465726,
1424
+ 2.736321,
1425
+ 2.704917,
1426
+ 4.216443,
1427
+ 4.26598
1428
+ ],
1429
+ "file": "task_058.pt"
1430
+ },
1431
+ {
1432
+ "index": 59,
1433
+ "dataset": "gas",
1434
+ "environment": "batch6",
1435
+ "task": "gas3_vs_5",
1436
+ "counts": {
1437
+ "source": 1276,
1438
+ "adaptation": 448,
1439
+ "evaluation": 268
1440
+ },
1441
+ "evaluation_hypothesis_counts": [
1442
+ 64,
1443
+ 204
1444
+ ],
1445
+ "group_counts": [
1446
+ 88,
1447
+ 30,
1448
+ 18
1449
+ ],
1450
+ "exclusions": [],
1451
+ "context_description": "provider temporal batch; source batches 1,2,3",
1452
+ "preprocessing_seconds": 0.0032335919968318194,
1453
+ "thresholds": [
1454
+ 4.68752,
1455
+ 4.364837,
1456
+ 4.216782,
1457
+ 4.26358,
1458
+ 2.711678,
1459
+ 2.642763,
1460
+ 4.47311,
1461
+ 4.454934
1462
+ ],
1463
+ "file": "task_059.pt"
1464
+ },
1465
+ {
1466
+ "index": 60,
1467
+ "dataset": "gas",
1468
+ "environment": "batch7",
1469
+ "task": "gas3_vs_5",
1470
+ "counts": {
1471
+ "source": 1276,
1472
+ "adaptation": 541,
1473
+ "evaluation": 449
1474
+ },
1475
+ "evaluation_hypothesis_counts": [
1476
+ 202,
1477
+ 247
1478
+ ],
1479
+ "group_counts": [
1480
+ 88,
1481
+ 72,
1482
+ 63
1483
+ ],
1484
+ "exclusions": [],
1485
+ "context_description": "provider temporal batch; source batches 1,2,3",
1486
+ "preprocessing_seconds": 0.0032499410153832287,
1487
+ "thresholds": [
1488
+ 4.68752,
1489
+ 4.364837,
1490
+ 4.216782,
1491
+ 4.26358,
1492
+ 2.711678,
1493
+ 2.642763,
1494
+ 4.47311,
1495
+ 4.454934
1496
+ ],
1497
+ "file": "task_060.pt"
1498
+ },
1499
+ {
1500
+ "index": 61,
1501
+ "dataset": "gas",
1502
+ "environment": "batch8",
1503
+ "task": "gas3_vs_5",
1504
+ "counts": {
1505
+ "source": 1276,
1506
+ "adaptation": 79,
1507
+ "evaluation": 104
1508
+ },
1509
+ "evaluation_hypothesis_counts": [
1510
+ 9,
1511
+ 95
1512
+ ],
1513
+ "group_counts": [
1514
+ 88,
1515
+ 6,
1516
+ 9
1517
+ ],
1518
+ "exclusions": [],
1519
+ "context_description": "provider temporal batch; source batches 1,2,3",
1520
+ "preprocessing_seconds": 0.0031235060014296323,
1521
+ "thresholds": [
1522
+ 4.68752,
1523
+ 4.364837,
1524
+ 4.216782,
1525
+ 4.26358,
1526
+ 2.711678,
1527
+ 2.642763,
1528
+ 4.47311,
1529
+ 4.454934
1530
+ ],
1531
+ "file": "task_061.pt"
1532
+ },
1533
+ {
1534
+ "index": 63,
1535
+ "dataset": "gas",
1536
+ "environment": "batch10",
1537
+ "task": "gas3_vs_5",
1538
+ "counts": {
1539
+ "source": 1276,
1540
+ "adaptation": 592,
1541
+ "evaluation": 608
1542
+ },
1543
+ "evaluation_hypothesis_counts": [
1544
+ 308,
1545
+ 300
1546
+ ],
1547
+ "group_counts": [
1548
+ 88,
1549
+ 75,
1550
+ 77
1551
+ ],
1552
+ "exclusions": [],
1553
+ "context_description": "provider temporal batch; source batches 1,2,3",
1554
+ "preprocessing_seconds": 0.0034391639928799123,
1555
+ "thresholds": [
1556
+ 4.68752,
1557
+ 4.364837,
1558
+ 4.216782,
1559
+ 4.26358,
1560
+ 2.711678,
1561
+ 2.642763,
1562
+ 4.47311,
1563
+ 4.454934
1564
+ ],
1565
+ "file": "task_063.pt"
1566
+ },
1567
+ {
1568
+ "index": 64,
1569
+ "dataset": "gas",
1570
+ "environment": "batch6",
1571
+ "task": "gas3_vs_6",
1572
+ "counts": {
1573
+ "source": 478,
1574
+ "adaptation": 333,
1575
+ "evaluation": 244
1576
+ },
1577
+ "evaluation_hypothesis_counts": [
1578
+ 64,
1579
+ 180
1580
+ ],
1581
+ "group_counts": [
1582
+ 34,
1583
+ 23,
1584
+ 16
1585
+ ],
1586
+ "exclusions": [],
1587
+ "context_description": "provider temporal batch; source batches 1,2,3",
1588
+ "preprocessing_seconds": 0.0031419199949596077,
1589
+ "thresholds": [
1590
+ 3.614317,
1591
+ 4.072812,
1592
+ 3.168788,
1593
+ 3.428528,
1594
+ 2.475792,
1595
+ 2.484679,
1596
+ 3.091101,
1597
+ 3.317349
1598
+ ],
1599
+ "file": "task_064.pt"
1600
+ },
1601
+ {
1602
+ "index": 65,
1603
+ "dataset": "gas",
1604
+ "environment": "batch7",
1605
+ "task": "gas3_vs_6",
1606
+ "counts": {
1607
+ "source": 478,
1608
+ "adaptation": 414,
1609
+ "evaluation": 514
1610
+ },
1611
+ "evaluation_hypothesis_counts": [
1612
+ 202,
1613
+ 312
1614
+ ],
1615
+ "group_counts": [
1616
+ 34,
1617
+ 49,
1618
+ 49
1619
+ ],
1620
+ "exclusions": [],
1621
+ "context_description": "provider temporal batch; source batches 1,2,3",
1622
+ "preprocessing_seconds": 0.0027873869985342026,
1623
+ "thresholds": [
1624
+ 3.614317,
1625
+ 4.072812,
1626
+ 3.168788,
1627
+ 3.428528,
1628
+ 2.475792,
1629
+ 2.484679,
1630
+ 3.091101,
1631
+ 3.317349
1632
+ ],
1633
+ "file": "task_065.pt"
1634
+ },
1635
+ {
1636
+ "index": 67,
1637
+ "dataset": "gas",
1638
+ "environment": "batch9",
1639
+ "task": "gas3_vs_6",
1640
+ "counts": {
1641
+ "source": 478,
1642
+ "adaptation": 104,
1643
+ "evaluation": 97
1644
+ },
1645
+ "evaluation_hypothesis_counts": [
1646
+ 48,
1647
+ 49
1648
+ ],
1649
+ "group_counts": [
1650
+ 34,
1651
+ 8,
1652
+ 7
1653
+ ],
1654
+ "exclusions": [],
1655
+ "context_description": "provider temporal batch; source batches 1,2,3",
1656
+ "preprocessing_seconds": 0.0029294709966052324,
1657
+ "thresholds": [
1658
+ 3.614317,
1659
+ 4.072812,
1660
+ 3.168788,
1661
+ 3.428528,
1662
+ 2.475792,
1663
+ 2.484679,
1664
+ 3.091101,
1665
+ 3.317349
1666
+ ],
1667
+ "file": "task_067.pt"
1668
+ },
1669
+ {
1670
+ "index": 68,
1671
+ "dataset": "gas",
1672
+ "environment": "batch10",
1673
+ "task": "gas3_vs_6",
1674
+ "counts": {
1675
+ "source": 478,
1676
+ "adaptation": 524,
1677
+ "evaluation": 676
1678
+ },
1679
+ "evaluation_hypothesis_counts": [
1680
+ 308,
1681
+ 368
1682
+ ],
1683
+ "group_counts": [
1684
+ 34,
1685
+ 54,
1686
+ 68
1687
+ ],
1688
+ "exclusions": [],
1689
+ "context_description": "provider temporal batch; source batches 1,2,3",
1690
+ "preprocessing_seconds": 0.0030813149933237582,
1691
+ "thresholds": [
1692
+ 3.614317,
1693
+ 4.072812,
1694
+ 3.168788,
1695
+ 3.428528,
1696
+ 2.475792,
1697
+ 2.484679,
1698
+ 3.091101,
1699
+ 3.317349
1700
+ ],
1701
+ "file": "task_068.pt"
1702
+ },
1703
+ {
1704
+ "index": 70,
1705
+ "dataset": "gas",
1706
+ "environment": "batch7",
1707
+ "task": "gas4_vs_5",
1708
+ "counts": {
1709
+ "source": 1256,
1710
+ "adaptation": 823,
1711
+ "evaluation": 551
1712
+ },
1713
+ "evaluation_hypothesis_counts": [
1714
+ 304,
1715
+ 247
1716
+ ],
1717
+ "group_counts": [
1718
+ 85,
1719
+ 74,
1720
+ 53
1721
+ ],
1722
+ "exclusions": [],
1723
+ "context_description": "provider temporal batch; source batches 1,2,3",
1724
+ "preprocessing_seconds": 0.0028891309921164066,
1725
+ "thresholds": [
1726
+ 6.378887,
1727
+ 5.284346,
1728
+ 4.988929,
1729
+ 4.757184,
1730
+ 2.77411,
1731
+ 2.68517,
1732
+ 5.359629,
1733
+ 5.33301
1734
+ ],
1735
+ "file": "task_070.pt"
1736
+ },
1737
+ {
1738
+ "index": 73,
1739
+ "dataset": "gas",
1740
+ "environment": "batch10",
1741
+ "task": "gas4_vs_5",
1742
+ "counts": {
1743
+ "source": 1256,
1744
+ "adaptation": 616,
1745
+ "evaluation": 584
1746
+ },
1747
+ "evaluation_hypothesis_counts": [
1748
+ 284,
1749
+ 300
1750
+ ],
1751
+ "group_counts": [
1752
+ 85,
1753
+ 63,
1754
+ 60
1755
+ ],
1756
+ "exclusions": [],
1757
+ "context_description": "provider temporal batch; source batches 1,2,3",
1758
+ "preprocessing_seconds": 0.0031111720018088818,
1759
+ "thresholds": [
1760
+ 6.378887,
1761
+ 5.284346,
1762
+ 4.988929,
1763
+ 4.757184,
1764
+ 2.77411,
1765
+ 2.68517,
1766
+ 5.359629,
1767
+ 5.33301
1768
+ ],
1769
+ "file": "task_073.pt"
1770
+ },
1771
+ {
1772
+ "index": 75,
1773
+ "dataset": "gas",
1774
+ "environment": "batch7",
1775
+ "task": "gas4_vs_6",
1776
+ "counts": {
1777
+ "source": 458,
1778
+ "adaptation": 696,
1779
+ "evaluation": 616
1780
+ },
1781
+ "evaluation_hypothesis_counts": [
1782
+ 304,
1783
+ 312
1784
+ ],
1785
+ "group_counts": [
1786
+ 33,
1787
+ 85,
1788
+ 74
1789
+ ],
1790
+ "exclusions": [],
1791
+ "context_description": "provider temporal batch; source batches 1,2,3",
1792
+ "preprocessing_seconds": 0.003027942992048338,
1793
+ "thresholds": [
1794
+ 10.822871,
1795
+ 9.340272,
1796
+ 5.247357,
1797
+ 5.060152,
1798
+ 2.990188,
1799
+ 2.945444,
1800
+ 6.379987,
1801
+ 6.278654
1802
+ ],
1803
+ "file": "task_075.pt"
1804
+ },
1805
+ {
1806
+ "index": 78,
1807
+ "dataset": "gas",
1808
+ "environment": "batch10",
1809
+ "task": "gas4_vs_6",
1810
+ "counts": {
1811
+ "source": 458,
1812
+ "adaptation": 548,
1813
+ "evaluation": 652
1814
+ },
1815
+ "evaluation_hypothesis_counts": [
1816
+ 284,
1817
+ 368
1818
+ ],
1819
+ "group_counts": [
1820
+ 33,
1821
+ 69,
1822
+ 83
1823
+ ],
1824
+ "exclusions": [],
1825
+ "context_description": "provider temporal batch; source batches 1,2,3",
1826
+ "preprocessing_seconds": 0.0030159790185280144,
1827
+ "thresholds": [
1828
+ 10.822871,
1829
+ 9.340272,
1830
+ 5.247357,
1831
+ 5.060152,
1832
+ 2.990188,
1833
+ 2.945444,
1834
+ 6.379987,
1835
+ 6.278654
1836
+ ],
1837
+ "file": "task_078.pt"
1838
+ },
1839
+ {
1840
+ "index": 79,
1841
+ "dataset": "gas",
1842
+ "environment": "batch6",
1843
+ "task": "gas5_vs_6",
1844
+ "counts": {
1845
+ "source": 956,
1846
+ "adaptation": 689,
1847
+ "evaluation": 384
1848
+ },
1849
+ "evaluation_hypothesis_counts": [
1850
+ 204,
1851
+ 180
1852
+ ],
1853
+ "group_counts": [
1854
+ 65,
1855
+ 45,
1856
+ 24
1857
+ ],
1858
+ "exclusions": [],
1859
+ "context_description": "provider temporal batch; source batches 1,2,3",
1860
+ "preprocessing_seconds": 0.0032675770053174347,
1861
+ "thresholds": [
1862
+ 5.626042,
1863
+ 4.708761,
1864
+ 4.907162,
1865
+ 4.690788,
1866
+ 2.754609,
1867
+ 2.67197,
1868
+ 5.239749,
1869
+ 5.190898
1870
+ ],
1871
+ "file": "task_079.pt"
1872
+ },
1873
+ {
1874
+ "index": 80,
1875
+ "dataset": "gas",
1876
+ "environment": "batch7",
1877
+ "task": "gas5_vs_6",
1878
+ "counts": {
1879
+ "source": 956,
1880
+ "adaptation": 639,
1881
+ "evaluation": 559
1882
+ },
1883
+ "evaluation_hypothesis_counts": [
1884
+ 247,
1885
+ 312
1886
+ ],
1887
+ "group_counts": [
1888
+ 65,
1889
+ 57,
1890
+ 52
1891
+ ],
1892
+ "exclusions": [],
1893
+ "context_description": "provider temporal batch; source batches 1,2,3",
1894
+ "preprocessing_seconds": 0.003296412993222475,
1895
+ "thresholds": [
1896
+ 5.626042,
1897
+ 4.708761,
1898
+ 4.907162,
1899
+ 4.690788,
1900
+ 2.754609,
1901
+ 2.67197,
1902
+ 5.239749,
1903
+ 5.190898
1904
+ ],
1905
+ "file": "task_080.pt"
1906
+ },
1907
+ {
1908
+ "index": 83,
1909
+ "dataset": "gas",
1910
+ "environment": "batch10",
1911
+ "task": "gas5_vs_6",
1912
+ "counts": {
1913
+ "source": 956,
1914
+ "adaptation": 532,
1915
+ "evaluation": 668
1916
+ },
1917
+ "evaluation_hypothesis_counts": [
1918
+ 300,
1919
+ 368
1920
+ ],
1921
+ "group_counts": [
1922
+ 65,
1923
+ 67,
1924
+ 85
1925
+ ],
1926
+ "exclusions": [],
1927
+ "context_description": "provider temporal batch; source batches 1,2,3",
1928
+ "preprocessing_seconds": 0.0031276129884645343,
1929
+ "thresholds": [
1930
+ 5.626042,
1931
+ 4.708761,
1932
+ 4.907162,
1933
+ 4.690788,
1934
+ 2.754609,
1935
+ 2.67197,
1936
+ 5.239749,
1937
+ 5.190898
1938
+ ],
1939
+ "file": "task_083.pt"
1940
+ }
1941
+ ],
1942
+ "excluded": [
1943
+ {
1944
+ "index": 11,
1945
+ "dataset": "gas",
1946
+ "environment": "batch8",
1947
+ "task": "gas1_vs_2",
1948
+ "counts": {
1949
+ "source": 1541,
1950
+ "adaptation": 38,
1951
+ "evaluation": 22
1952
+ },
1953
+ "evaluation_hypothesis_counts": [
1954
+ 11,
1955
+ 11
1956
+ ],
1957
+ "group_counts": [
1958
+ 101,
1959
+ 3,
1960
+ 2
1961
+ ],
1962
+ "exclusions": [
1963
+ "adaptation_support",
1964
+ "evaluation_support",
1965
+ "evaluation_group_support"
1966
+ ],
1967
+ "context_description": "provider temporal batch; source batches 1,2,3"
1968
+ },
1969
+ {
1970
+ "index": 12,
1971
+ "dataset": "gas",
1972
+ "environment": "batch9",
1973
+ "task": "gas1_vs_2",
1974
+ "counts": {
1975
+ "source": 1541,
1976
+ "adaptation": 58,
1977
+ "evaluation": 58
1978
+ },
1979
+ "evaluation_hypothesis_counts": [
1980
+ 42,
1981
+ 16
1982
+ ],
1983
+ "group_counts": [
1984
+ 101,
1985
+ 4,
1986
+ 4
1987
+ ],
1988
+ "exclusions": [
1989
+ "adaptation_support"
1990
+ ],
1991
+ "context_description": "provider temporal batch; source batches 1,2,3"
1992
+ },
1993
+ {
1994
+ "index": 16,
1995
+ "dataset": "gas",
1996
+ "environment": "batch8",
1997
+ "task": "gas1_vs_3",
1998
+ "counts": {
1999
+ "source": 1018,
2000
+ "adaptation": 50,
2001
+ "evaluation": 20
2002
+ },
2003
+ "evaluation_hypothesis_counts": [
2004
+ 11,
2005
+ 9
2006
+ ],
2007
+ "group_counts": [
2008
+ 70,
2009
+ 6,
2010
+ 4
2011
+ ],
2012
+ "exclusions": [
2013
+ "adaptation_support",
2014
+ "evaluation_support"
2015
+ ],
2016
+ "context_description": "provider temporal batch; source batches 1,2,3"
2017
+ },
2018
+ {
2019
+ "index": 19,
2020
+ "dataset": "gas",
2021
+ "environment": "batch6",
2022
+ "task": "gas1_vs_4",
2023
+ "counts": {
2024
+ "source": 998,
2025
+ "adaptation": 266,
2026
+ "evaluation": 277
2027
+ },
2028
+ "evaluation_hypothesis_counts": [
2029
+ 277,
2030
+ 0
2031
+ ],
2032
+ "group_counts": [
2033
+ 71,
2034
+ 17,
2035
+ 18
2036
+ ],
2037
+ "exclusions": [
2038
+ "evaluation_hypothesis_support"
2039
+ ],
2040
+ "context_description": "provider temporal batch; source batches 1,2,3"
2041
+ },
2042
+ {
2043
+ "index": 21,
2044
+ "dataset": "gas",
2045
+ "environment": "batch8",
2046
+ "task": "gas1_vs_4",
2047
+ "counts": {
2048
+ "source": 998,
2049
+ "adaptation": 28,
2050
+ "evaluation": 35
2051
+ },
2052
+ "evaluation_hypothesis_counts": [
2053
+ 11,
2054
+ 24
2055
+ ],
2056
+ "group_counts": [
2057
+ 71,
2058
+ 5,
2059
+ 7
2060
+ ],
2061
+ "exclusions": [
2062
+ "adaptation_support"
2063
+ ],
2064
+ "context_description": "provider temporal batch; source batches 1,2,3"
2065
+ },
2066
+ {
2067
+ "index": 22,
2068
+ "dataset": "gas",
2069
+ "environment": "batch9",
2070
+ "task": "gas1_vs_4",
2071
+ "counts": {
2072
+ "source": 998,
2073
+ "adaptation": 29,
2074
+ "evaluation": 107
2075
+ },
2076
+ "evaluation_hypothesis_counts": [
2077
+ 42,
2078
+ 65
2079
+ ],
2080
+ "group_counts": [
2081
+ 71,
2082
+ 4,
2083
+ 8
2084
+ ],
2085
+ "exclusions": [
2086
+ "adaptation_support"
2087
+ ],
2088
+ "context_description": "provider temporal batch; source batches 1,2,3"
2089
+ },
2090
+ {
2091
+ "index": 27,
2092
+ "dataset": "gas",
2093
+ "environment": "batch9",
2094
+ "task": "gas1_vs_5",
2095
+ "counts": {
2096
+ "source": 1496,
2097
+ "adaptation": 29,
2098
+ "evaluation": 110
2099
+ },
2100
+ "evaluation_hypothesis_counts": [
2101
+ 42,
2102
+ 68
2103
+ ],
2104
+ "group_counts": [
2105
+ 103,
2106
+ 4,
2107
+ 8
2108
+ ],
2109
+ "exclusions": [
2110
+ "adaptation_support"
2111
+ ],
2112
+ "context_description": "provider temporal batch; source batches 1,2,3"
2113
+ },
2114
+ {
2115
+ "index": 31,
2116
+ "dataset": "gas",
2117
+ "environment": "batch8",
2118
+ "task": "gas1_vs_6",
2119
+ "counts": {
2120
+ "source": 698,
2121
+ "adaptation": 27,
2122
+ "evaluation": 21
2123
+ },
2124
+ "evaluation_hypothesis_counts": [
2125
+ 11,
2126
+ 10
2127
+ ],
2128
+ "group_counts": [
2129
+ 50,
2130
+ 4,
2131
+ 3
2132
+ ],
2133
+ "exclusions": [
2134
+ "adaptation_support",
2135
+ "evaluation_support",
2136
+ "evaluation_group_support"
2137
+ ],
2138
+ "context_description": "provider temporal batch; source batches 1,2,3"
2139
+ },
2140
+ {
2141
+ "index": 36,
2142
+ "dataset": "gas",
2143
+ "environment": "batch8",
2144
+ "task": "gas2_vs_3",
2145
+ "counts": {
2146
+ "source": 1321,
2147
+ "adaptation": 50,
2148
+ "evaluation": 20
2149
+ },
2150
+ "evaluation_hypothesis_counts": [
2151
+ 11,
2152
+ 9
2153
+ ],
2154
+ "group_counts": [
2155
+ 89,
2156
+ 6,
2157
+ 4
2158
+ ],
2159
+ "exclusions": [
2160
+ "adaptation_support",
2161
+ "evaluation_support"
2162
+ ],
2163
+ "context_description": "provider temporal batch; source batches 1,2,3"
2164
+ },
2165
+ {
2166
+ "index": 39,
2167
+ "dataset": "gas",
2168
+ "environment": "batch6",
2169
+ "task": "gas2_vs_4",
2170
+ "counts": {
2171
+ "source": 1301,
2172
+ "adaptation": 400,
2173
+ "evaluation": 203
2174
+ },
2175
+ "evaluation_hypothesis_counts": [
2176
+ 203,
2177
+ 0
2178
+ ],
2179
+ "group_counts": [
2180
+ 90,
2181
+ 26,
2182
+ 15
2183
+ ],
2184
+ "exclusions": [
2185
+ "evaluation_hypothesis_support"
2186
+ ],
2187
+ "context_description": "provider temporal batch; source batches 1,2,3"
2188
+ },
2189
+ {
2190
+ "index": 41,
2191
+ "dataset": "gas",
2192
+ "environment": "batch8",
2193
+ "task": "gas2_vs_4",
2194
+ "counts": {
2195
+ "source": 1301,
2196
+ "adaptation": 28,
2197
+ "evaluation": 35
2198
+ },
2199
+ "evaluation_hypothesis_counts": [
2200
+ 11,
2201
+ 24
2202
+ ],
2203
+ "group_counts": [
2204
+ 90,
2205
+ 5,
2206
+ 7
2207
+ ],
2208
+ "exclusions": [
2209
+ "adaptation_support"
2210
+ ],
2211
+ "context_description": "provider temporal batch; source batches 1,2,3"
2212
+ },
2213
+ {
2214
+ "index": 42,
2215
+ "dataset": "gas",
2216
+ "environment": "batch9",
2217
+ "task": "gas2_vs_4",
2218
+ "counts": {
2219
+ "source": 1301,
2220
+ "adaptation": 49,
2221
+ "evaluation": 81
2222
+ },
2223
+ "evaluation_hypothesis_counts": [
2224
+ 16,
2225
+ 65
2226
+ ],
2227
+ "group_counts": [
2228
+ 90,
2229
+ 5,
2230
+ 7
2231
+ ],
2232
+ "exclusions": [
2233
+ "adaptation_support"
2234
+ ],
2235
+ "context_description": "provider temporal batch; source batches 1,2,3"
2236
+ },
2237
+ {
2238
+ "index": 47,
2239
+ "dataset": "gas",
2240
+ "environment": "batch9",
2241
+ "task": "gas2_vs_5",
2242
+ "counts": {
2243
+ "source": 1799,
2244
+ "adaptation": 49,
2245
+ "evaluation": 84
2246
+ },
2247
+ "evaluation_hypothesis_counts": [
2248
+ 16,
2249
+ 68
2250
+ ],
2251
+ "group_counts": [
2252
+ 125,
2253
+ 5,
2254
+ 7
2255
+ ],
2256
+ "exclusions": [
2257
+ "adaptation_support"
2258
+ ],
2259
+ "context_description": "provider temporal batch; source batches 1,2,3"
2260
+ },
2261
+ {
2262
+ "index": 51,
2263
+ "dataset": "gas",
2264
+ "environment": "batch8",
2265
+ "task": "gas2_vs_6",
2266
+ "counts": {
2267
+ "source": 1001,
2268
+ "adaptation": 27,
2269
+ "evaluation": 21
2270
+ },
2271
+ "evaluation_hypothesis_counts": [
2272
+ 11,
2273
+ 10
2274
+ ],
2275
+ "group_counts": [
2276
+ 71,
2277
+ 4,
2278
+ 3
2279
+ ],
2280
+ "exclusions": [
2281
+ "adaptation_support",
2282
+ "evaluation_support",
2283
+ "evaluation_group_support"
2284
+ ],
2285
+ "context_description": "provider temporal batch; source batches 1,2,3"
2286
+ },
2287
+ {
2288
+ "index": 54,
2289
+ "dataset": "gas",
2290
+ "environment": "batch6",
2291
+ "task": "gas3_vs_4",
2292
+ "counts": {
2293
+ "source": 778,
2294
+ "adaptation": 75,
2295
+ "evaluation": 64
2296
+ },
2297
+ "evaluation_hypothesis_counts": [
2298
+ 64,
2299
+ 0
2300
+ ],
2301
+ "group_counts": [
2302
+ 53,
2303
+ 5,
2304
+ 4
2305
+ ],
2306
+ "exclusions": [
2307
+ "evaluation_hypothesis_support"
2308
+ ],
2309
+ "context_description": "provider temporal batch; source batches 1,2,3"
2310
+ },
2311
+ {
2312
+ "index": 56,
2313
+ "dataset": "gas",
2314
+ "environment": "batch8",
2315
+ "task": "gas3_vs_4",
2316
+ "counts": {
2317
+ "source": 778,
2318
+ "adaptation": 40,
2319
+ "evaluation": 33
2320
+ },
2321
+ "evaluation_hypothesis_counts": [
2322
+ 9,
2323
+ 24
2324
+ ],
2325
+ "group_counts": [
2326
+ 53,
2327
+ 4,
2328
+ 6
2329
+ ],
2330
+ "exclusions": [
2331
+ "adaptation_support"
2332
+ ],
2333
+ "context_description": "provider temporal batch; source batches 1,2,3"
2334
+ },
2335
+ {
2336
+ "index": 57,
2337
+ "dataset": "gas",
2338
+ "environment": "batch9",
2339
+ "task": "gas3_vs_4",
2340
+ "counts": {
2341
+ "source": 778,
2342
+ "adaptation": 62,
2343
+ "evaluation": 113
2344
+ },
2345
+ "evaluation_hypothesis_counts": [
2346
+ 48,
2347
+ 65
2348
+ ],
2349
+ "group_counts": [
2350
+ 53,
2351
+ 5,
2352
+ 9
2353
+ ],
2354
+ "exclusions": [
2355
+ "adaptation_support"
2356
+ ],
2357
+ "context_description": "provider temporal batch; source batches 1,2,3"
2358
+ },
2359
+ {
2360
+ "index": 62,
2361
+ "dataset": "gas",
2362
+ "environment": "batch9",
2363
+ "task": "gas3_vs_5",
2364
+ "counts": {
2365
+ "source": 1276,
2366
+ "adaptation": 62,
2367
+ "evaluation": 116
2368
+ },
2369
+ "evaluation_hypothesis_counts": [
2370
+ 48,
2371
+ 68
2372
+ ],
2373
+ "group_counts": [
2374
+ 88,
2375
+ 6,
2376
+ 9
2377
+ ],
2378
+ "exclusions": [
2379
+ "adaptation_support"
2380
+ ],
2381
+ "context_description": "provider temporal batch; source batches 1,2,3"
2382
+ },
2383
+ {
2384
+ "index": 66,
2385
+ "dataset": "gas",
2386
+ "environment": "batch8",
2387
+ "task": "gas3_vs_6",
2388
+ "counts": {
2389
+ "source": 478,
2390
+ "adaptation": 39,
2391
+ "evaluation": 19
2392
+ },
2393
+ "evaluation_hypothesis_counts": [
2394
+ 9,
2395
+ 10
2396
+ ],
2397
+ "group_counts": [
2398
+ 34,
2399
+ 3,
2400
+ 3
2401
+ ],
2402
+ "exclusions": [
2403
+ "adaptation_support",
2404
+ "evaluation_support",
2405
+ "evaluation_group_support"
2406
+ ],
2407
+ "context_description": "provider temporal batch; source batches 1,2,3"
2408
+ },
2409
+ {
2410
+ "index": 69,
2411
+ "dataset": "gas",
2412
+ "environment": "batch6",
2413
+ "task": "gas4_vs_5",
2414
+ "counts": {
2415
+ "source": 1256,
2416
+ "adaptation": 431,
2417
+ "evaluation": 204
2418
+ },
2419
+ "evaluation_hypothesis_counts": [
2420
+ 0,
2421
+ 204
2422
+ ],
2423
+ "group_counts": [
2424
+ 85,
2425
+ 27,
2426
+ 14
2427
+ ],
2428
+ "exclusions": [
2429
+ "evaluation_hypothesis_support"
2430
+ ],
2431
+ "context_description": "provider temporal batch; source batches 1,2,3"
2432
+ },
2433
+ {
2434
+ "index": 71,
2435
+ "dataset": "gas",
2436
+ "environment": "batch8",
2437
+ "task": "gas4_vs_5",
2438
+ "counts": {
2439
+ "source": 1256,
2440
+ "adaptation": 57,
2441
+ "evaluation": 119
2442
+ },
2443
+ "evaluation_hypothesis_counts": [
2444
+ 24,
2445
+ 95
2446
+ ],
2447
+ "group_counts": [
2448
+ 85,
2449
+ 5,
2450
+ 8
2451
+ ],
2452
+ "exclusions": [
2453
+ "adaptation_support"
2454
+ ],
2455
+ "context_description": "provider temporal batch; source batches 1,2,3"
2456
+ },
2457
+ {
2458
+ "index": 72,
2459
+ "dataset": "gas",
2460
+ "environment": "batch9",
2461
+ "task": "gas4_vs_5",
2462
+ "counts": {
2463
+ "source": 1256,
2464
+ "adaptation": 20,
2465
+ "evaluation": 133
2466
+ },
2467
+ "evaluation_hypothesis_counts": [
2468
+ 65,
2469
+ 68
2470
+ ],
2471
+ "group_counts": [
2472
+ 85,
2473
+ 3,
2474
+ 9
2475
+ ],
2476
+ "exclusions": [
2477
+ "adaptation_support"
2478
+ ],
2479
+ "context_description": "provider temporal batch; source batches 1,2,3"
2480
+ },
2481
+ {
2482
+ "index": 74,
2483
+ "dataset": "gas",
2484
+ "environment": "batch6",
2485
+ "task": "gas4_vs_6",
2486
+ "counts": {
2487
+ "source": 458,
2488
+ "adaptation": 316,
2489
+ "evaluation": 180
2490
+ },
2491
+ "evaluation_hypothesis_counts": [
2492
+ 0,
2493
+ 180
2494
+ ],
2495
+ "group_counts": [
2496
+ 33,
2497
+ 21,
2498
+ 12
2499
+ ],
2500
+ "exclusions": [
2501
+ "evaluation_hypothesis_support"
2502
+ ],
2503
+ "context_description": "provider temporal batch; source batches 1,2,3"
2504
+ },
2505
+ {
2506
+ "index": 76,
2507
+ "dataset": "gas",
2508
+ "environment": "batch8",
2509
+ "task": "gas4_vs_6",
2510
+ "counts": {
2511
+ "source": 458,
2512
+ "adaptation": 17,
2513
+ "evaluation": 34
2514
+ },
2515
+ "evaluation_hypothesis_counts": [
2516
+ 24,
2517
+ 10
2518
+ ],
2519
+ "group_counts": [
2520
+ 33,
2521
+ 3,
2522
+ 6
2523
+ ],
2524
+ "exclusions": [
2525
+ "adaptation_support"
2526
+ ],
2527
+ "context_description": "provider temporal batch; source batches 1,2,3"
2528
+ },
2529
+ {
2530
+ "index": 77,
2531
+ "dataset": "gas",
2532
+ "environment": "batch9",
2533
+ "task": "gas4_vs_6",
2534
+ "counts": {
2535
+ "source": 458,
2536
+ "adaptation": 62,
2537
+ "evaluation": 114
2538
+ },
2539
+ "evaluation_hypothesis_counts": [
2540
+ 65,
2541
+ 49
2542
+ ],
2543
+ "group_counts": [
2544
+ 33,
2545
+ 6,
2546
+ 9
2547
+ ],
2548
+ "exclusions": [
2549
+ "adaptation_support"
2550
+ ],
2551
+ "context_description": "provider temporal batch; source batches 1,2,3"
2552
+ },
2553
+ {
2554
+ "index": 81,
2555
+ "dataset": "gas",
2556
+ "environment": "batch8",
2557
+ "task": "gas5_vs_6",
2558
+ "counts": {
2559
+ "source": 956,
2560
+ "adaptation": 56,
2561
+ "evaluation": 105
2562
+ },
2563
+ "evaluation_hypothesis_counts": [
2564
+ 95,
2565
+ 10
2566
+ ],
2567
+ "group_counts": [
2568
+ 65,
2569
+ 5,
2570
+ 9
2571
+ ],
2572
+ "exclusions": [
2573
+ "adaptation_support"
2574
+ ],
2575
+ "context_description": "provider temporal batch; source batches 1,2,3"
2576
+ },
2577
+ {
2578
+ "index": 82,
2579
+ "dataset": "gas",
2580
+ "environment": "batch9",
2581
+ "task": "gas5_vs_6",
2582
+ "counts": {
2583
+ "source": 956,
2584
+ "adaptation": 62,
2585
+ "evaluation": 117
2586
+ },
2587
+ "evaluation_hypothesis_counts": [
2588
+ 68,
2589
+ 49
2590
+ ],
2591
+ "group_counts": [
2592
+ 65,
2593
+ 5,
2594
+ 9
2595
+ ],
2596
+ "exclusions": [
2597
+ "adaptation_support"
2598
+ ],
2599
+ "context_description": "provider temporal batch; source batches 1,2,3"
2600
+ }
2601
+ ],
2602
+ "total_declared_tasks": 84,
2603
+ "raw_parse_and_aggregation_seconds": 8.298483947000932,
2604
+ "fixed_binary_adapter": true,
2605
+ "provider_defined_sequential_task": false,
2606
+ "target_configuration_or_contiguous_block_leakage": false,
2607
+ "pipeline_note": "All recorded future bits are evaluator/archival payloads. Policy construction sees only adaptation/source samples."
2608
+ }
provenance/hydraulic/confirmation_lock.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "unionmind/continuous_decision/classifiers.py": "ebb1bfcd9d72469368ffc14cbb847faeb53a69d19abc9b684546f3791d61e20b",
4
+ "unionmind/continuous_decision/__init__.py": "3a95b4d2e04edd607dbf129eb72b173f8addbc66086484720edea2b0e4e09d76",
5
+ "unionmind/continuous_decision/factor_acquisition.py": "771f560388ba9a24adadd7fd84d5bb90ff5f52e16ffc92e121a0aa4ba0a91b51",
6
+ "unionmind/continuous_decision/metric.py": "74919a6b6e5142da90cf63f3a154426ee195c699c744e34ce9207253544b2023",
7
+ "unionmind/continuous_decision/mixture.py": "2d25dfbc9df8cb9b471c247c88913fc9f91bc9aa6bc5c8d8e4e412fd7965612e",
8
+ "unionmind/continuous_decision/factor_study.py": "beb76b2cc09aaa2b796e8bf69f0235fb9b0c33823ce1e33ea8170d56b60ba109",
9
+ "unionmind/continuous_decision/learned_controls.py": "81e3419c8f20422ed495fab8b43fa264dceb8503d1a3ad71cdc70b4c32b0d6f3",
10
+ "unionmind/continuous_decision/acquisition_analysis.py": "75bd00b2c5374c4ad11588285a92d589509e8798b9c432bd5c0484aa54dcd6a1",
11
+ "unionmind/continuous_decision/factor_analysis.py": "0fb267e3c0572db48f4a2ca41fc4777ec082039c8312f66a75005355837c3fb8",
12
+ "unionmind/continuous_decision/interval_value.py": "cd3c0158a377c97fa9f4dc2f56c2a9597840a68bb07d37782b42265f727e4fe9",
13
+ "unionmind/continuous_decision/factorial.py": "54241b65b32042937346369414d750fe58042f0eae20f25c391a74a480e065f0",
14
+ "unionmind/continuous_decision/source.py": "1c011b7608637c73652224573caf152cefc3297c2f94599b8f88816ded829601",
15
+ "unionmind/continuous_decision/gaussian.py": "588a3e5f39a79fe26b959b7b2a62c6700bdd242e03a6c2adbd674f8d6537a2e5",
16
+ "configs/factorial_transfer_protocol_01.json": "aadd4b386cd447cdf8e4fbb19f8925b2a73ee86006dc5ab45e298d4a864c7d7c",
17
+ "tests/test_continuous_decision.py": "b5e0f4e887f8483b41b56e179b10bf9349d5a18367eb47f5634c98a8ffb023ce"
18
+ },
19
+ "timestamp_utc": "2026-10-20T12:55:21Z",
20
+ "before_new_gas_archive_or_numerical_values": true,
21
+ "hydraulic_is_opened_development": true,
22
+ "candidate_and_all_thresholds_fixed": true
23
+ }
provenance/hydraulic/opened_hydraulic_summary.json ADDED
@@ -0,0 +1,695 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "rows": [
3
+ {
4
+ "samples": 64,
5
+ "model": "additive_full",
6
+ "records": 304,
7
+ "groups": 83,
8
+ "top3": 0.33442982456140347,
9
+ "top1": 0.09758771929824561,
10
+ "seed_top3": [
11
+ 0.21052631578947367,
12
+ 0.3223684210526316,
13
+ 0.47039473684210525
14
+ ]
15
+ },
16
+ {
17
+ "samples": 64,
18
+ "model": "additive_diag",
19
+ "records": 304,
20
+ "groups": 83,
21
+ "top3": 0.17105263157894735,
22
+ "top1": 0.07236842105263158,
23
+ "seed_top3": [
24
+ 0.09868421052631579,
25
+ 0.1644736842105263,
26
+ 0.25
27
+ ]
28
+ },
29
+ {
30
+ "samples": 64,
31
+ "model": "factorized_neural",
32
+ "records": 304,
33
+ "groups": 83,
34
+ "top3": 0.46820175438596484,
35
+ "top1": 0.19956140350877188,
36
+ "seed_top3": [
37
+ 0.49671052631578944,
38
+ 0.5361842105263157,
39
+ 0.37171052631578944
40
+ ]
41
+ },
42
+ {
43
+ "samples": 64,
44
+ "model": "joint_neural",
45
+ "records": 304,
46
+ "groups": 83,
47
+ "top3": 0.002295147926284236,
48
+ "top1": 0.0,
49
+ "seed_top3": [
50
+ 0.002982827832292596,
51
+ 0.003704113224237063,
52
+ 0.00019850272232304898
53
+ ]
54
+ },
55
+ {
56
+ "samples": 64,
57
+ "model": "factorized_nb",
58
+ "records": 304,
59
+ "groups": 83,
60
+ "top3": 0.03357403653919592,
61
+ "top1": 0.010964912280701754,
62
+ "seed_top3": [
63
+ 2.3165307635285396e-05,
64
+ 0.05712532463829721,
65
+ 0.04357361967165525
66
+ ]
67
+ },
68
+ {
69
+ "samples": 64,
70
+ "model": "factorized_lda",
71
+ "records": 304,
72
+ "groups": 83,
73
+ "top3": 0.16008771929824558,
74
+ "top1": 0.06578947368421051,
75
+ "seed_top3": [
76
+ 0.1282894736842105,
77
+ 0.12171052631578946,
78
+ 0.23026315789473684
79
+ ]
80
+ },
81
+ {
82
+ "samples": 64,
83
+ "model": "factorized_kernel",
84
+ "records": 304,
85
+ "groups": 83,
86
+ "top3": 0.10307017543859648,
87
+ "top1": 0.014254385964912278,
88
+ "seed_top3": [
89
+ 0.10526315789473684,
90
+ 0.046052631578947366,
91
+ 0.15789473684210525
92
+ ]
93
+ },
94
+ {
95
+ "samples": 64,
96
+ "model": "factorized_ridge",
97
+ "records": 304,
98
+ "groups": 83,
99
+ "top3": 0.16118421052631576,
100
+ "top1": 0.03728070175438596,
101
+ "seed_top3": [
102
+ 0.15460526315789472,
103
+ 0.09868421052631579,
104
+ 0.23026315789473684
105
+ ]
106
+ },
107
+ {
108
+ "samples": 64,
109
+ "model": "joint_gaussian",
110
+ "records": 304,
111
+ "groups": 83,
112
+ "top3": 0.0002856405720182809,
113
+ "top1": 0.0,
114
+ "seed_top3": [
115
+ 0.0,
116
+ 0.000856921716054843,
117
+ 0.0
118
+ ]
119
+ },
120
+ {
121
+ "samples": 64,
122
+ "model": "factorized_logistic",
123
+ "records": 304,
124
+ "groups": 83,
125
+ "top3": 0.18311403508771928,
126
+ "top1": 0.04824561403508771,
127
+ "seed_top3": [
128
+ 0.17434210526315788,
129
+ 0.19407894736842105,
130
+ 0.18092105263157893
131
+ ]
132
+ },
133
+ {
134
+ "samples": 64,
135
+ "model": "factorized_classifier",
136
+ "records": 304,
137
+ "groups": 83,
138
+ "top3": 0.2554824561403508,
139
+ "top1": 0.07456140350877191,
140
+ "seed_top3": [
141
+ 0.24013157894736842,
142
+ 0.24013157894736842,
143
+ 0.28618421052631576
144
+ ]
145
+ },
146
+ {
147
+ "samples": 128,
148
+ "model": "additive_full",
149
+ "records": 304,
150
+ "groups": 83,
151
+ "top3": 0.48355263157894735,
152
+ "top1": 0.16666666666666666,
153
+ "seed_top3": [
154
+ 0.4375,
155
+ 0.5164473684210527,
156
+ 0.49671052631578944
157
+ ]
158
+ },
159
+ {
160
+ "samples": 128,
161
+ "model": "additive_diag",
162
+ "records": 304,
163
+ "groups": 83,
164
+ "top3": 0.22258771929824558,
165
+ "top1": 0.07017543859649122,
166
+ "seed_top3": [
167
+ 0.3421052631578947,
168
+ 0.1381578947368421,
169
+ 0.1875
170
+ ]
171
+ },
172
+ {
173
+ "samples": 128,
174
+ "model": "factorized_neural",
175
+ "records": 304,
176
+ "groups": 83,
177
+ "top3": 0.32675438596491224,
178
+ "top1": 0.1019736842105263,
179
+ "seed_top3": [
180
+ 0.3881578947368421,
181
+ 0.23684210526315788,
182
+ 0.35526315789473684
183
+ ]
184
+ },
185
+ {
186
+ "samples": 128,
187
+ "model": "joint_neural",
188
+ "records": 304,
189
+ "groups": 83,
190
+ "top3": 0.0012024117093776534,
191
+ "top1": 1.0749914000687993e-05,
192
+ "seed_top3": [
193
+ 0.0019809911450778326,
194
+ 0.0016262439830551285,
195
+ 0.0
196
+ ]
197
+ },
198
+ {
199
+ "samples": 128,
200
+ "model": "factorized_nb",
201
+ "records": 304,
202
+ "groups": 83,
203
+ "top3": 0.046168458117123785,
204
+ "top1": 0.0,
205
+ "seed_top3": [
206
+ 0.03319588584136397,
207
+ 0.06910211267605634,
208
+ 0.03620737583395107
209
+ ]
210
+ },
211
+ {
212
+ "samples": 128,
213
+ "model": "factorized_lda",
214
+ "records": 304,
215
+ "groups": 83,
216
+ "top3": 0.17214912280701752,
217
+ "top1": 0.04714912280701754,
218
+ "seed_top3": [
219
+ 0.15789473684210525,
220
+ 0.1381578947368421,
221
+ 0.22039473684210525
222
+ ]
223
+ },
224
+ {
225
+ "samples": 128,
226
+ "model": "factorized_kernel",
227
+ "records": 304,
228
+ "groups": 83,
229
+ "top3": 0.15685958778064038,
230
+ "top1": 0.025219298245614034,
231
+ "seed_top3": [
232
+ 0.22373021715126976,
233
+ 0.11855907250644092,
234
+ 0.1282894736842105
235
+ ]
236
+ },
237
+ {
238
+ "samples": 128,
239
+ "model": "factorized_ridge",
240
+ "records": 304,
241
+ "groups": 83,
242
+ "top3": 0.20833333333333331,
243
+ "top1": 0.04934210526315789,
244
+ "seed_top3": [
245
+ 0.20723684210526314,
246
+ 0.1875,
247
+ 0.23026315789473684
248
+ ]
249
+ },
250
+ {
251
+ "samples": 128,
252
+ "model": "joint_gaussian",
253
+ "records": 304,
254
+ "groups": 83,
255
+ "top3": 3.0671083302662245e-05,
256
+ "top1": 0.0,
257
+ "seed_top3": [
258
+ 4.6006624953993374e-05,
259
+ 4.6006624953993374e-05,
260
+ 0.0
261
+ ]
262
+ },
263
+ {
264
+ "samples": 128,
265
+ "model": "factorized_logistic",
266
+ "records": 304,
267
+ "groups": 83,
268
+ "top3": 0.35197368421052627,
269
+ "top1": 0.12609649122807015,
270
+ "seed_top3": [
271
+ 0.29605263157894735,
272
+ 0.4013157894736842,
273
+ 0.35855263157894735
274
+ ]
275
+ },
276
+ {
277
+ "samples": 128,
278
+ "model": "factorized_classifier",
279
+ "records": 304,
280
+ "groups": 83,
281
+ "top3": 0.5219298245614035,
282
+ "top1": 0.23684210526315788,
283
+ "seed_top3": [
284
+ 0.5197368421052632,
285
+ 0.6282894736842105,
286
+ 0.41776315789473684
287
+ ]
288
+ }
289
+ ],
290
+ "comparisons": [
291
+ {
292
+ "samples": 64,
293
+ "candidate": "additive_full",
294
+ "control": "additive_diag",
295
+ "gain": 0.16337719298245612,
296
+ "simultaneous_interval": [
297
+ 0.08264551530341001,
298
+ 0.25458812531422864
299
+ ]
300
+ },
301
+ {
302
+ "samples": 64,
303
+ "candidate": "additive_full",
304
+ "control": "factorized_neural",
305
+ "gain": -0.1337719298245615,
306
+ "simultaneous_interval": [
307
+ -0.2761466773680033,
308
+ 0.013853212771188953
309
+ ]
310
+ },
311
+ {
312
+ "samples": 64,
313
+ "candidate": "additive_full",
314
+ "control": "joint_neural",
315
+ "gain": 0.33213467663511925,
316
+ "simultaneous_interval": [
317
+ 0.22690524278499632,
318
+ 0.4407765636597409
319
+ ]
320
+ },
321
+ {
322
+ "samples": 64,
323
+ "candidate": "additive_full",
324
+ "control": "factorized_nb",
325
+ "gain": 0.3008557880222075,
326
+ "simultaneous_interval": [
327
+ 0.20067828169201005,
328
+ 0.4049301748867331
329
+ ]
330
+ },
331
+ {
332
+ "samples": 64,
333
+ "candidate": "additive_full",
334
+ "control": "factorized_lda",
335
+ "gain": 0.17434210526315783,
336
+ "simultaneous_interval": [
337
+ 0.08102048432981543,
338
+ 0.27345139127606827
339
+ ]
340
+ },
341
+ {
342
+ "samples": 64,
343
+ "candidate": "additive_full",
344
+ "control": "factorized_kernel",
345
+ "gain": 0.231359649122807,
346
+ "simultaneous_interval": [
347
+ 0.1306394912131519,
348
+ 0.33820839884696097
349
+ ]
350
+ },
351
+ {
352
+ "samples": 64,
353
+ "candidate": "additive_full",
354
+ "control": "factorized_ridge",
355
+ "gain": 0.17324561403508767,
356
+ "simultaneous_interval": [
357
+ 0.0791346581350637,
358
+ 0.2787748640935527
359
+ ]
360
+ },
361
+ {
362
+ "samples": 64,
363
+ "candidate": "additive_full",
364
+ "control": "joint_gaussian",
365
+ "gain": 0.33414418398938517,
366
+ "simultaneous_interval": [
367
+ 0.2289311196644366,
368
+ 0.44346739389679396
369
+ ]
370
+ },
371
+ {
372
+ "samples": 64,
373
+ "candidate": "additive_full",
374
+ "control": "factorized_logistic",
375
+ "gain": 0.15131578947368415,
376
+ "simultaneous_interval": [
377
+ 0.06250062542566985,
378
+ 0.24781685327230032
379
+ ]
380
+ },
381
+ {
382
+ "samples": 64,
383
+ "candidate": "additive_full",
384
+ "control": "factorized_classifier",
385
+ "gain": 0.0789473684210526,
386
+ "simultaneous_interval": [
387
+ -0.019466464058081168,
388
+ 0.17948968879938865
389
+ ]
390
+ },
391
+ {
392
+ "samples": 64,
393
+ "candidate": "factorized_neural",
394
+ "control": "additive_full",
395
+ "gain": 0.1337719298245615,
396
+ "simultaneous_interval": [
397
+ -0.013853212771187994,
398
+ 0.2761466773680043
399
+ ]
400
+ },
401
+ {
402
+ "samples": 64,
403
+ "candidate": "factorized_neural",
404
+ "control": "additive_diag",
405
+ "gain": 0.2971491228070176,
406
+ "simultaneous_interval": [
407
+ 0.15426113387978133,
408
+ 0.44105000258478105
409
+ ]
410
+ },
411
+ {
412
+ "samples": 64,
413
+ "candidate": "factorized_neural",
414
+ "control": "joint_neural",
415
+ "gain": 0.46590660645968074,
416
+ "simultaneous_interval": [
417
+ 0.35046650479566105,
418
+ 0.5893573556971095
419
+ ]
420
+ },
421
+ {
422
+ "samples": 64,
423
+ "candidate": "factorized_neural",
424
+ "control": "factorized_nb",
425
+ "gain": 0.434627717846769,
426
+ "simultaneous_interval": [
427
+ 0.3040597064119879,
428
+ 0.5743779242108623
429
+ ]
430
+ },
431
+ {
432
+ "samples": 64,
433
+ "candidate": "factorized_neural",
434
+ "control": "factorized_lda",
435
+ "gain": 0.3081140350877193,
436
+ "simultaneous_interval": [
437
+ 0.14866567800648,
438
+ 0.4681410326086966
439
+ ]
440
+ },
441
+ {
442
+ "samples": 64,
443
+ "candidate": "factorized_neural",
444
+ "control": "factorized_kernel",
445
+ "gain": 0.3651315789473685,
446
+ "simultaneous_interval": [
447
+ 0.233542714884696,
448
+ 0.49620915471923605
449
+ ]
450
+ },
451
+ {
452
+ "samples": 64,
453
+ "candidate": "factorized_neural",
454
+ "control": "factorized_ridge",
455
+ "gain": 0.3070175438596492,
456
+ "simultaneous_interval": [
457
+ 0.16909983593900785,
458
+ 0.4482647898587899
459
+ ]
460
+ },
461
+ {
462
+ "samples": 64,
463
+ "candidate": "factorized_neural",
464
+ "control": "joint_gaussian",
465
+ "gain": 0.46791611381394665,
466
+ "simultaneous_interval": [
467
+ 0.35256662331044836,
468
+ 0.5916442864271282
469
+ ]
470
+ },
471
+ {
472
+ "samples": 64,
473
+ "candidate": "factorized_neural",
474
+ "control": "factorized_logistic",
475
+ "gain": 0.2850877192982456,
476
+ "simultaneous_interval": [
477
+ 0.15062503327583265,
478
+ 0.4259581612178132
479
+ ]
480
+ },
481
+ {
482
+ "samples": 64,
483
+ "candidate": "factorized_neural",
484
+ "control": "factorized_classifier",
485
+ "gain": 0.21271929824561409,
486
+ "simultaneous_interval": [
487
+ 0.07962219374376149,
488
+ 0.35669512433095546
489
+ ]
490
+ },
491
+ {
492
+ "samples": 128,
493
+ "candidate": "additive_full",
494
+ "control": "additive_diag",
495
+ "gain": 0.26096491228070184,
496
+ "simultaneous_interval": [
497
+ 0.17175747863247864,
498
+ 0.35717504216711177
499
+ ]
500
+ },
501
+ {
502
+ "samples": 128,
503
+ "candidate": "additive_full",
504
+ "control": "factorized_neural",
505
+ "gain": 0.15679824561403516,
506
+ "simultaneous_interval": [
507
+ 0.018288378916756023,
508
+ 0.2977022177605471
509
+ ]
510
+ },
511
+ {
512
+ "samples": 128,
513
+ "candidate": "additive_full",
514
+ "control": "joint_neural",
515
+ "gain": 0.48235021986956983,
516
+ "simultaneous_interval": [
517
+ 0.36131032544628366,
518
+ 0.6059796551907133
519
+ ]
520
+ },
521
+ {
522
+ "samples": 128,
523
+ "candidate": "additive_full",
524
+ "control": "factorized_nb",
525
+ "gain": 0.4373841734618237,
526
+ "simultaneous_interval": [
527
+ 0.29736921445546155,
528
+ 0.5767076143558851
529
+ ]
530
+ },
531
+ {
532
+ "samples": 128,
533
+ "candidate": "additive_full",
534
+ "control": "factorized_lda",
535
+ "gain": 0.3114035087719299,
536
+ "simultaneous_interval": [
537
+ 0.16611773574561403,
538
+ 0.456033862999231
539
+ ]
540
+ },
541
+ {
542
+ "samples": 128,
543
+ "candidate": "additive_full",
544
+ "control": "factorized_kernel",
545
+ "gain": 0.3266930437983071,
546
+ "simultaneous_interval": [
547
+ 0.1790342981683134,
548
+ 0.47683126571354323
549
+ ]
550
+ },
551
+ {
552
+ "samples": 128,
553
+ "candidate": "additive_full",
554
+ "control": "factorized_ridge",
555
+ "gain": 0.27521929824561414,
556
+ "simultaneous_interval": [
557
+ 0.1535565803749844,
558
+ 0.3977193713450295
559
+ ]
560
+ },
561
+ {
562
+ "samples": 128,
563
+ "candidate": "additive_full",
564
+ "control": "joint_gaussian",
565
+ "gain": 0.4835219604956448,
566
+ "simultaneous_interval": [
567
+ 0.3628264719751996,
568
+ 0.6070391413265996
569
+ ]
570
+ },
571
+ {
572
+ "samples": 128,
573
+ "candidate": "additive_full",
574
+ "control": "factorized_logistic",
575
+ "gain": 0.13157894736842113,
576
+ "simultaneous_interval": [
577
+ 0.0023228803716609236,
578
+ 0.2539563715269826
579
+ ]
580
+ },
581
+ {
582
+ "samples": 128,
583
+ "candidate": "additive_full",
584
+ "control": "factorized_classifier",
585
+ "gain": -0.03837719298245601,
586
+ "simultaneous_interval": [
587
+ -0.14913320543927824,
588
+ 0.07941876576397329
589
+ ]
590
+ },
591
+ {
592
+ "samples": 128,
593
+ "candidate": "factorized_neural",
594
+ "control": "additive_full",
595
+ "gain": -0.15679824561403516,
596
+ "simultaneous_interval": [
597
+ -0.29770221776054645,
598
+ -0.01828837891675554
599
+ ]
600
+ },
601
+ {
602
+ "samples": 128,
603
+ "candidate": "factorized_neural",
604
+ "control": "additive_diag",
605
+ "gain": 0.10416666666666671,
606
+ "simultaneous_interval": [
607
+ -0.04301137368755138,
608
+ 0.257489592625633
609
+ ]
610
+ },
611
+ {
612
+ "samples": 128,
613
+ "candidate": "factorized_neural",
614
+ "control": "joint_neural",
615
+ "gain": 0.32555197425553467,
616
+ "simultaneous_interval": [
617
+ 0.20589475912776967,
618
+ 0.45345039636467305
619
+ ]
620
+ },
621
+ {
622
+ "samples": 128,
623
+ "candidate": "factorized_neural",
624
+ "control": "factorized_nb",
625
+ "gain": 0.2805859278477885,
626
+ "simultaneous_interval": [
627
+ 0.15287488074668454,
628
+ 0.41681037577302665
629
+ ]
630
+ },
631
+ {
632
+ "samples": 128,
633
+ "candidate": "factorized_neural",
634
+ "control": "factorized_lda",
635
+ "gain": 0.15460526315789477,
636
+ "simultaneous_interval": [
637
+ 0.0005317129629630314,
638
+ 0.310613476335657
639
+ ]
640
+ },
641
+ {
642
+ "samples": 128,
643
+ "candidate": "factorized_neural",
644
+ "control": "factorized_kernel",
645
+ "gain": 0.16989479818427192,
646
+ "simultaneous_interval": [
647
+ 0.024059249947207276,
648
+ 0.31291910635660686
649
+ ]
650
+ },
651
+ {
652
+ "samples": 128,
653
+ "candidate": "factorized_neural",
654
+ "control": "factorized_ridge",
655
+ "gain": 0.11842105263157898,
656
+ "simultaneous_interval": [
657
+ -0.0287217138973585,
658
+ 0.2756078250509257
659
+ ]
660
+ },
661
+ {
662
+ "samples": 128,
663
+ "candidate": "factorized_neural",
664
+ "control": "joint_gaussian",
665
+ "gain": 0.3267237148816096,
666
+ "simultaneous_interval": [
667
+ 0.20691513932731084,
668
+ 0.4544342401229345
669
+ ]
670
+ },
671
+ {
672
+ "samples": 128,
673
+ "candidate": "factorized_neural",
674
+ "control": "factorized_logistic",
675
+ "gain": -0.02521929824561403,
676
+ "simultaneous_interval": [
677
+ -0.18147220133353387,
678
+ 0.12980928807423497
679
+ ]
680
+ },
681
+ {
682
+ "samples": 128,
683
+ "candidate": "factorized_neural",
684
+ "control": "factorized_classifier",
685
+ "gain": -0.19517543859649117,
686
+ "simultaneous_interval": [
687
+ -0.343900221378874,
688
+ -0.04644709422751629
689
+ ]
690
+ }
691
+ ],
692
+ "exploratory_opened_source": true,
693
+ "no_confirmation_claim": true,
694
+ "seconds": 0.3426705059973756
695
+ }
provenance/temperature/acquisition_summary.json ADDED
@@ -0,0 +1,678 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "rows": [
3
+ {
4
+ "model": "additive_full",
5
+ "policy": "value_forced",
6
+ "budget": 2,
7
+ "top3": 0.3125,
8
+ "queries": 2.0,
9
+ "recorded_feature_bytes": 16.0,
10
+ "complete_episode_seconds": 2.4728256590024102,
11
+ "exact_value_evaluations": 13440
12
+ },
13
+ {
14
+ "model": "additive_full",
15
+ "policy": "value_forced",
16
+ "budget": 4,
17
+ "top3": 0.34375,
18
+ "queries": 4.0,
19
+ "recorded_feature_bytes": 32.0,
20
+ "complete_episode_seconds": 2.4728256590024102,
21
+ "exact_value_evaluations": 13440
22
+ },
23
+ {
24
+ "model": "additive_full",
25
+ "policy": "value_forced",
26
+ "budget": 8,
27
+ "top3": 0.34375,
28
+ "queries": 8.0,
29
+ "recorded_feature_bytes": 64.0,
30
+ "complete_episode_seconds": 2.4728256590024102,
31
+ "exact_value_evaluations": 13440
32
+ },
33
+ {
34
+ "model": "additive_full",
35
+ "policy": "value_stop",
36
+ "budget": 2,
37
+ "top3": 0.3125,
38
+ "queries": 2.0,
39
+ "recorded_feature_bytes": 16.0,
40
+ "complete_episode_seconds": 2.103017521003494,
41
+ "exact_value_evaluations": 12245
42
+ },
43
+ {
44
+ "model": "additive_full",
45
+ "policy": "value_stop",
46
+ "budget": 4,
47
+ "top3": 0.34375,
48
+ "queries": 3.9375,
49
+ "recorded_feature_bytes": 31.5,
50
+ "complete_episode_seconds": 2.103017521003494,
51
+ "exact_value_evaluations": 12245
52
+ },
53
+ {
54
+ "model": "additive_full",
55
+ "policy": "value_stop",
56
+ "budget": 8,
57
+ "top3": 0.34375,
58
+ "queries": 6.71875,
59
+ "recorded_feature_bytes": 53.75,
60
+ "complete_episode_seconds": 2.103017521003494,
61
+ "exact_value_evaluations": 12245
62
+ },
63
+ {
64
+ "model": "additive_full",
65
+ "policy": "variance_forced",
66
+ "budget": 2,
67
+ "top3": 0.3125,
68
+ "queries": 2.0,
69
+ "recorded_feature_bytes": 16.0,
70
+ "complete_episode_seconds": 0.11508204501296859,
71
+ "exact_value_evaluations": 0
72
+ },
73
+ {
74
+ "model": "additive_full",
75
+ "policy": "variance_forced",
76
+ "budget": 4,
77
+ "top3": 0.3125,
78
+ "queries": 4.0,
79
+ "recorded_feature_bytes": 32.0,
80
+ "complete_episode_seconds": 0.11508204501296859,
81
+ "exact_value_evaluations": 0
82
+ },
83
+ {
84
+ "model": "additive_full",
85
+ "policy": "variance_forced",
86
+ "budget": 8,
87
+ "top3": 0.375,
88
+ "queries": 8.0,
89
+ "recorded_feature_bytes": 64.0,
90
+ "complete_episode_seconds": 0.11508204501296859,
91
+ "exact_value_evaluations": 0
92
+ },
93
+ {
94
+ "model": "additive_full",
95
+ "policy": "variance_stop",
96
+ "budget": 2,
97
+ "top3": 0.3125,
98
+ "queries": 2.0,
99
+ "recorded_feature_bytes": 16.0,
100
+ "complete_episode_seconds": 0.15590574299858417,
101
+ "exact_value_evaluations": 207
102
+ },
103
+ {
104
+ "model": "additive_full",
105
+ "policy": "variance_stop",
106
+ "budget": 4,
107
+ "top3": 0.3125,
108
+ "queries": 3.78125,
109
+ "recorded_feature_bytes": 30.25,
110
+ "complete_episode_seconds": 0.15590574299858417,
111
+ "exact_value_evaluations": 207
112
+ },
113
+ {
114
+ "model": "additive_full",
115
+ "policy": "variance_stop",
116
+ "budget": 8,
117
+ "top3": 0.34375,
118
+ "queries": 5.6875,
119
+ "recorded_feature_bytes": 45.5,
120
+ "complete_episode_seconds": 0.15590574299858417,
121
+ "exact_value_evaluations": 207
122
+ },
123
+ {
124
+ "model": "additive_full",
125
+ "policy": "static_forced",
126
+ "budget": 2,
127
+ "top3": 0.28125,
128
+ "queries": 2.0,
129
+ "recorded_feature_bytes": 16.0,
130
+ "complete_episode_seconds": 0.010446411994053051,
131
+ "exact_value_evaluations": 0
132
+ },
133
+ {
134
+ "model": "additive_full",
135
+ "policy": "static_forced",
136
+ "budget": 4,
137
+ "top3": 0.28125,
138
+ "queries": 4.0,
139
+ "recorded_feature_bytes": 32.0,
140
+ "complete_episode_seconds": 0.010446411994053051,
141
+ "exact_value_evaluations": 0
142
+ },
143
+ {
144
+ "model": "additive_full",
145
+ "policy": "static_forced",
146
+ "budget": 8,
147
+ "top3": 0.28125,
148
+ "queries": 8.0,
149
+ "recorded_feature_bytes": 64.0,
150
+ "complete_episode_seconds": 0.010446411994053051,
151
+ "exact_value_evaluations": 0
152
+ },
153
+ {
154
+ "model": "additive_full",
155
+ "policy": "random_forced",
156
+ "budget": 2,
157
+ "top3": 0.15625,
158
+ "queries": 2.0,
159
+ "recorded_feature_bytes": 16.0,
160
+ "complete_episode_seconds": 0.20825171600154135,
161
+ "exact_value_evaluations": 0
162
+ },
163
+ {
164
+ "model": "additive_full",
165
+ "policy": "random_forced",
166
+ "budget": 4,
167
+ "top3": 0.1875,
168
+ "queries": 4.0,
169
+ "recorded_feature_bytes": 32.0,
170
+ "complete_episode_seconds": 0.20825171600154135,
171
+ "exact_value_evaluations": 0
172
+ },
173
+ {
174
+ "model": "additive_full",
175
+ "policy": "random_forced",
176
+ "budget": 8,
177
+ "top3": 0.25,
178
+ "queries": 8.0,
179
+ "recorded_feature_bytes": 64.0,
180
+ "complete_episode_seconds": 0.20825171600154135,
181
+ "exact_value_evaluations": 0
182
+ },
183
+ {
184
+ "model": "factorized_neural",
185
+ "policy": "value_forced",
186
+ "budget": 2,
187
+ "top3": 0.6875,
188
+ "queries": 2.0,
189
+ "recorded_feature_bytes": 16.0,
190
+ "complete_episode_seconds": 2.471479789004661,
191
+ "exact_value_evaluations": 13440
192
+ },
193
+ {
194
+ "model": "factorized_neural",
195
+ "policy": "value_forced",
196
+ "budget": 4,
197
+ "top3": 0.59375,
198
+ "queries": 4.0,
199
+ "recorded_feature_bytes": 32.0,
200
+ "complete_episode_seconds": 2.471479789004661,
201
+ "exact_value_evaluations": 13440
202
+ },
203
+ {
204
+ "model": "factorized_neural",
205
+ "policy": "value_forced",
206
+ "budget": 8,
207
+ "top3": 0.59375,
208
+ "queries": 8.0,
209
+ "recorded_feature_bytes": 64.0,
210
+ "complete_episode_seconds": 2.471479789004661,
211
+ "exact_value_evaluations": 13440
212
+ },
213
+ {
214
+ "model": "factorized_neural",
215
+ "policy": "value_stop",
216
+ "budget": 2,
217
+ "top3": 0.6875,
218
+ "queries": 2.0,
219
+ "recorded_feature_bytes": 16.0,
220
+ "complete_episode_seconds": 1.7633266810007626,
221
+ "exact_value_evaluations": 11170
222
+ },
223
+ {
224
+ "model": "factorized_neural",
225
+ "policy": "value_stop",
226
+ "budget": 4,
227
+ "top3": 0.59375,
228
+ "queries": 3.65625,
229
+ "recorded_feature_bytes": 29.25,
230
+ "complete_episode_seconds": 1.7633266810007626,
231
+ "exact_value_evaluations": 11170
232
+ },
233
+ {
234
+ "model": "factorized_neural",
235
+ "policy": "value_stop",
236
+ "budget": 8,
237
+ "top3": 0.59375,
238
+ "queries": 6.0,
239
+ "recorded_feature_bytes": 48.0,
240
+ "complete_episode_seconds": 1.7633266810007626,
241
+ "exact_value_evaluations": 11170
242
+ },
243
+ {
244
+ "model": "factorized_neural",
245
+ "policy": "variance_forced",
246
+ "budget": 2,
247
+ "top3": 0.65625,
248
+ "queries": 2.0,
249
+ "recorded_feature_bytes": 16.0,
250
+ "complete_episode_seconds": 0.10477982299926225,
251
+ "exact_value_evaluations": 0
252
+ },
253
+ {
254
+ "model": "factorized_neural",
255
+ "policy": "variance_forced",
256
+ "budget": 4,
257
+ "top3": 0.59375,
258
+ "queries": 4.0,
259
+ "recorded_feature_bytes": 32.0,
260
+ "complete_episode_seconds": 0.10477982299926225,
261
+ "exact_value_evaluations": 0
262
+ },
263
+ {
264
+ "model": "factorized_neural",
265
+ "policy": "variance_forced",
266
+ "budget": 8,
267
+ "top3": 0.59375,
268
+ "queries": 8.0,
269
+ "recorded_feature_bytes": 64.0,
270
+ "complete_episode_seconds": 0.10477982299926225,
271
+ "exact_value_evaluations": 0
272
+ },
273
+ {
274
+ "model": "factorized_neural",
275
+ "policy": "variance_stop",
276
+ "budget": 2,
277
+ "top3": 0.65625,
278
+ "queries": 2.0,
279
+ "recorded_feature_bytes": 16.0,
280
+ "complete_episode_seconds": 0.13838808199216146,
281
+ "exact_value_evaluations": 206
282
+ },
283
+ {
284
+ "model": "factorized_neural",
285
+ "policy": "variance_stop",
286
+ "budget": 4,
287
+ "top3": 0.59375,
288
+ "queries": 3.78125,
289
+ "recorded_feature_bytes": 30.25,
290
+ "complete_episode_seconds": 0.13838808199216146,
291
+ "exact_value_evaluations": 206
292
+ },
293
+ {
294
+ "model": "factorized_neural",
295
+ "policy": "variance_stop",
296
+ "budget": 8,
297
+ "top3": 0.59375,
298
+ "queries": 5.8125,
299
+ "recorded_feature_bytes": 46.5,
300
+ "complete_episode_seconds": 0.13838808199216146,
301
+ "exact_value_evaluations": 206
302
+ },
303
+ {
304
+ "model": "factorized_neural",
305
+ "policy": "static_forced",
306
+ "budget": 2,
307
+ "top3": 0.6875,
308
+ "queries": 2.0,
309
+ "recorded_feature_bytes": 16.0,
310
+ "complete_episode_seconds": 0.01028569700429216,
311
+ "exact_value_evaluations": 0
312
+ },
313
+ {
314
+ "model": "factorized_neural",
315
+ "policy": "static_forced",
316
+ "budget": 4,
317
+ "top3": 0.65625,
318
+ "queries": 4.0,
319
+ "recorded_feature_bytes": 32.0,
320
+ "complete_episode_seconds": 0.01028569700429216,
321
+ "exact_value_evaluations": 0
322
+ },
323
+ {
324
+ "model": "factorized_neural",
325
+ "policy": "static_forced",
326
+ "budget": 8,
327
+ "top3": 0.6875,
328
+ "queries": 8.0,
329
+ "recorded_feature_bytes": 64.0,
330
+ "complete_episode_seconds": 0.01028569700429216,
331
+ "exact_value_evaluations": 0
332
+ },
333
+ {
334
+ "model": "factorized_neural",
335
+ "policy": "random_forced",
336
+ "budget": 2,
337
+ "top3": 0.34375,
338
+ "queries": 2.0,
339
+ "recorded_feature_bytes": 16.0,
340
+ "complete_episode_seconds": 0.2074921750026988,
341
+ "exact_value_evaluations": 0
342
+ },
343
+ {
344
+ "model": "factorized_neural",
345
+ "policy": "random_forced",
346
+ "budget": 4,
347
+ "top3": 0.40625,
348
+ "queries": 4.0,
349
+ "recorded_feature_bytes": 32.0,
350
+ "complete_episode_seconds": 0.2074921750026988,
351
+ "exact_value_evaluations": 0
352
+ },
353
+ {
354
+ "model": "factorized_neural",
355
+ "policy": "random_forced",
356
+ "budget": 8,
357
+ "top3": 0.5625,
358
+ "queries": 8.0,
359
+ "recorded_feature_bytes": 64.0,
360
+ "complete_episode_seconds": 0.2074921750026988,
361
+ "exact_value_evaluations": 0
362
+ }
363
+ ],
364
+ "contrasts": [
365
+ {
366
+ "model": "additive_full",
367
+ "budget": 2,
368
+ "control": "variance_forced",
369
+ "gain": 0.0,
370
+ "simultaneous_interval": [
371
+ 0.0,
372
+ 0.0
373
+ ]
374
+ },
375
+ {
376
+ "model": "additive_full",
377
+ "budget": 2,
378
+ "control": "variance_stop",
379
+ "gain": 0.0,
380
+ "simultaneous_interval": [
381
+ 0.0,
382
+ 0.0
383
+ ]
384
+ },
385
+ {
386
+ "model": "additive_full",
387
+ "budget": 2,
388
+ "control": "static_forced",
389
+ "gain": 0.03125,
390
+ "simultaneous_interval": [
391
+ 0.0,
392
+ 0.15625
393
+ ]
394
+ },
395
+ {
396
+ "model": "additive_full",
397
+ "budget": 2,
398
+ "control": "random_forced",
399
+ "gain": 0.15625,
400
+ "simultaneous_interval": [
401
+ -0.15625,
402
+ 0.46875
403
+ ]
404
+ },
405
+ {
406
+ "model": "additive_full",
407
+ "budget": 4,
408
+ "control": "variance_forced",
409
+ "gain": 0.03125,
410
+ "simultaneous_interval": [
411
+ 0.0,
412
+ 0.15625
413
+ ]
414
+ },
415
+ {
416
+ "model": "additive_full",
417
+ "budget": 4,
418
+ "control": "variance_stop",
419
+ "gain": 0.03125,
420
+ "simultaneous_interval": [
421
+ 0.0,
422
+ 0.15625
423
+ ]
424
+ },
425
+ {
426
+ "model": "additive_full",
427
+ "budget": 4,
428
+ "control": "static_forced",
429
+ "gain": 0.0625,
430
+ "simultaneous_interval": [
431
+ 0.0,
432
+ 0.25
433
+ ]
434
+ },
435
+ {
436
+ "model": "additive_full",
437
+ "budget": 4,
438
+ "control": "random_forced",
439
+ "gain": 0.15625,
440
+ "simultaneous_interval": [
441
+ 0.0,
442
+ 0.375
443
+ ]
444
+ },
445
+ {
446
+ "model": "additive_full",
447
+ "budget": 8,
448
+ "control": "variance_forced",
449
+ "gain": -0.03125,
450
+ "simultaneous_interval": [
451
+ -0.15625,
452
+ 0.0
453
+ ]
454
+ },
455
+ {
456
+ "model": "additive_full",
457
+ "budget": 8,
458
+ "control": "variance_stop",
459
+ "gain": 0.0,
460
+ "simultaneous_interval": [
461
+ 0.0,
462
+ 0.0
463
+ ]
464
+ },
465
+ {
466
+ "model": "additive_full",
467
+ "budget": 8,
468
+ "control": "static_forced",
469
+ "gain": 0.0625,
470
+ "simultaneous_interval": [
471
+ 0.0,
472
+ 0.25
473
+ ]
474
+ },
475
+ {
476
+ "model": "additive_full",
477
+ "budget": 8,
478
+ "control": "random_forced",
479
+ "gain": 0.09375,
480
+ "simultaneous_interval": [
481
+ -0.125,
482
+ 0.34375
483
+ ]
484
+ },
485
+ {
486
+ "model": "factorized_neural",
487
+ "budget": 2,
488
+ "control": "variance_forced",
489
+ "gain": 0.03125,
490
+ "simultaneous_interval": [
491
+ -0.125,
492
+ 0.1875
493
+ ]
494
+ },
495
+ {
496
+ "model": "factorized_neural",
497
+ "budget": 2,
498
+ "control": "variance_stop",
499
+ "gain": 0.03125,
500
+ "simultaneous_interval": [
501
+ -0.125,
502
+ 0.1875
503
+ ]
504
+ },
505
+ {
506
+ "model": "factorized_neural",
507
+ "budget": 2,
508
+ "control": "static_forced",
509
+ "gain": 0.0,
510
+ "simultaneous_interval": [
511
+ -0.28125,
512
+ 0.28125
513
+ ]
514
+ },
515
+ {
516
+ "model": "factorized_neural",
517
+ "budget": 2,
518
+ "control": "random_forced",
519
+ "gain": 0.34375,
520
+ "simultaneous_interval": [
521
+ 0.0625,
522
+ 0.625
523
+ ]
524
+ },
525
+ {
526
+ "model": "factorized_neural",
527
+ "budget": 4,
528
+ "control": "variance_forced",
529
+ "gain": 0.0,
530
+ "simultaneous_interval": [
531
+ 0.0,
532
+ 0.0
533
+ ]
534
+ },
535
+ {
536
+ "model": "factorized_neural",
537
+ "budget": 4,
538
+ "control": "variance_stop",
539
+ "gain": 0.0,
540
+ "simultaneous_interval": [
541
+ 0.0,
542
+ 0.0
543
+ ]
544
+ },
545
+ {
546
+ "model": "factorized_neural",
547
+ "budget": 4,
548
+ "control": "static_forced",
549
+ "gain": -0.0625,
550
+ "simultaneous_interval": [
551
+ -0.40625,
552
+ 0.28125
553
+ ]
554
+ },
555
+ {
556
+ "model": "factorized_neural",
557
+ "budget": 4,
558
+ "control": "random_forced",
559
+ "gain": 0.1875,
560
+ "simultaneous_interval": [
561
+ -0.15625,
562
+ 0.5
563
+ ]
564
+ },
565
+ {
566
+ "model": "factorized_neural",
567
+ "budget": 8,
568
+ "control": "variance_forced",
569
+ "gain": 0.0,
570
+ "simultaneous_interval": [
571
+ 0.0,
572
+ 0.0
573
+ ]
574
+ },
575
+ {
576
+ "model": "factorized_neural",
577
+ "budget": 8,
578
+ "control": "variance_stop",
579
+ "gain": 0.0,
580
+ "simultaneous_interval": [
581
+ 0.0,
582
+ 0.0
583
+ ]
584
+ },
585
+ {
586
+ "model": "factorized_neural",
587
+ "budget": 8,
588
+ "control": "static_forced",
589
+ "gain": -0.09375,
590
+ "simultaneous_interval": [
591
+ -0.40625,
592
+ 0.21875
593
+ ]
594
+ },
595
+ {
596
+ "model": "factorized_neural",
597
+ "budget": 8,
598
+ "control": "random_forced",
599
+ "gain": 0.03125,
600
+ "simultaneous_interval": [
601
+ -0.1875,
602
+ 0.28125
603
+ ]
604
+ }
605
+ ],
606
+ "gates": [
607
+ {
608
+ "model": "additive_full",
609
+ "budget": 2,
610
+ "substantial_query_policy_gain": false,
611
+ "promotion": false
612
+ },
613
+ {
614
+ "model": "additive_full",
615
+ "budget": 4,
616
+ "substantial_query_policy_gain": false,
617
+ "promotion": false
618
+ },
619
+ {
620
+ "model": "additive_full",
621
+ "budget": 8,
622
+ "substantial_query_policy_gain": false,
623
+ "promotion": false
624
+ },
625
+ {
626
+ "model": "factorized_neural",
627
+ "budget": 2,
628
+ "substantial_query_policy_gain": false,
629
+ "promotion": false
630
+ },
631
+ {
632
+ "model": "factorized_neural",
633
+ "budget": 4,
634
+ "substantial_query_policy_gain": false,
635
+ "promotion": false
636
+ },
637
+ {
638
+ "model": "factorized_neural",
639
+ "budget": 8,
640
+ "substantial_query_policy_gain": false,
641
+ "promotion": false
642
+ }
643
+ ],
644
+ "full_references": [
645
+ {
646
+ "model": "additive_full",
647
+ "records": 32,
648
+ "queries": 56,
649
+ "top3": 0.3125,
650
+ "all_channels_paid": true
651
+ },
652
+ {
653
+ "model": "factorized_neural",
654
+ "records": 32,
655
+ "queries": 56,
656
+ "top3": 0.625,
657
+ "all_channels_paid": true
658
+ },
659
+ {
660
+ "model": "factorized_lda",
661
+ "records": 32,
662
+ "queries": 56,
663
+ "top3": 0.65625,
664
+ "all_channels_paid": true
665
+ },
666
+ {
667
+ "model": "factorized_classifier",
668
+ "records": 32,
669
+ "queries": 56,
670
+ "top3": 0.40625,
671
+ "all_channels_paid": true
672
+ }
673
+ ],
674
+ "identical_preselected_32_episodes": true,
675
+ "selection_rule_unchanged": true,
676
+ "physical_transport_bytes_not_measured": true,
677
+ "no_secondary_rescue_of_primary": true
678
+ }
provenance/temperature/cost_and_replay.json ADDED
@@ -0,0 +1,1006 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "python_import_and_cuda_startup_seconds": 18.972809764003614,
3
+ "cold_process_after_startup_seconds": 1.0000426979968324,
4
+ "inference_repetitions": 10,
5
+ "device": "cuda:2",
6
+ "gpu": "NVIDIA H200",
7
+ "peak_gpu_bytes": 112525824,
8
+ "results": [
9
+ {
10
+ "samples": 64,
11
+ "seed": 20851,
12
+ "model": "additive_full",
13
+ "held_records": 404,
14
+ "complete_load_and_normalization_seconds": 0.006550619989866391,
15
+ "inference_mean_seconds": 0.0002888571980292909,
16
+ "inference_p95_seconds": 0.00040566524694440865,
17
+ "model_unique_tensor_bytes": 45520,
18
+ "checkpoint_file_bytes": 50120,
19
+ "neural_parameters": 0,
20
+ "adaptation_seconds": 0.17665776300418656,
21
+ "replay_identical": true,
22
+ "all_adaptation_examples_source_only": true
23
+ },
24
+ {
25
+ "samples": 64,
26
+ "seed": 20851,
27
+ "model": "additive_diag",
28
+ "held_records": 404,
29
+ "complete_load_and_normalization_seconds": 0.0011089440085925162,
30
+ "inference_mean_seconds": 0.0002642215957166627,
31
+ "inference_p95_seconds": 0.0002991964967804961,
32
+ "model_unique_tensor_bytes": 45520,
33
+ "checkpoint_file_bytes": 50120,
34
+ "neural_parameters": 0,
35
+ "adaptation_seconds": 0.001148302006185986,
36
+ "replay_identical": true,
37
+ "all_adaptation_examples_source_only": true
38
+ },
39
+ {
40
+ "samples": 64,
41
+ "seed": 20851,
42
+ "model": "factorized_neural",
43
+ "held_records": 404,
44
+ "complete_load_and_normalization_seconds": 0.0012703450047411025,
45
+ "inference_mean_seconds": 0.00026501620159251614,
46
+ "inference_p95_seconds": 0.0002900718536693602,
47
+ "model_unique_tensor_bytes": 118992,
48
+ "checkpoint_file_bytes": 124555,
49
+ "neural_parameters": 19936,
50
+ "adaptation_seconds": 2.436637737991987,
51
+ "replay_identical": true,
52
+ "all_adaptation_examples_source_only": true
53
+ },
54
+ {
55
+ "samples": 64,
56
+ "seed": 20851,
57
+ "model": "joint_neural",
58
+ "held_records": 404,
59
+ "complete_load_and_normalization_seconds": 0.001231597998412326,
60
+ "inference_mean_seconds": 0.00026190730277448894,
61
+ "inference_p95_seconds": 0.0002831609555869363,
62
+ "model_unique_tensor_bytes": 119164,
63
+ "checkpoint_file_bytes": 124672,
64
+ "neural_parameters": 19979,
65
+ "adaptation_seconds": 0.8920095740031684,
66
+ "replay_identical": true,
67
+ "all_adaptation_examples_source_only": true
68
+ },
69
+ {
70
+ "samples": 64,
71
+ "seed": 20851,
72
+ "model": "factorized_nb",
73
+ "held_records": 404,
74
+ "complete_load_and_normalization_seconds": 0.001397985004587099,
75
+ "inference_mean_seconds": 0.0006190863001393154,
76
+ "inference_p95_seconds": 0.0006891197474033105,
77
+ "model_unique_tensor_bytes": 362248,
78
+ "checkpoint_file_bytes": 367954,
79
+ "neural_parameters": 0,
80
+ "adaptation_seconds": 0.0010508430132176727,
81
+ "replay_identical": true,
82
+ "all_adaptation_examples_source_only": true
83
+ },
84
+ {
85
+ "samples": 64,
86
+ "seed": 20851,
87
+ "model": "factorized_lda",
88
+ "held_records": 404,
89
+ "complete_load_and_normalization_seconds": 0.0012515370035544038,
90
+ "inference_mean_seconds": 0.0006096810990129598,
91
+ "inference_p95_seconds": 0.0006509788989205844,
92
+ "model_unique_tensor_bytes": 86280,
93
+ "checkpoint_file_bytes": 92002,
94
+ "neural_parameters": 0,
95
+ "adaptation_seconds": 0.0009277429926441982,
96
+ "replay_identical": true,
97
+ "all_adaptation_examples_source_only": true
98
+ },
99
+ {
100
+ "samples": 64,
101
+ "seed": 20851,
102
+ "model": "factorized_kernel",
103
+ "held_records": 404,
104
+ "complete_load_and_normalization_seconds": 0.0012337929947534576,
105
+ "inference_mean_seconds": 0.002624167398607824,
106
+ "inference_p95_seconds": 0.0027149345558427737,
107
+ "model_unique_tensor_bytes": 86280,
108
+ "checkpoint_file_bytes": 92050,
109
+ "neural_parameters": 0,
110
+ "adaptation_seconds": 0.0009275009942939505,
111
+ "replay_identical": true,
112
+ "all_adaptation_examples_source_only": true
113
+ },
114
+ {
115
+ "samples": 64,
116
+ "seed": 20851,
117
+ "model": "factorized_ridge",
118
+ "held_records": 404,
119
+ "complete_load_and_normalization_seconds": 0.0013018590107094496,
120
+ "inference_mean_seconds": 0.00025459540192969144,
121
+ "inference_p95_seconds": 0.0003088367477175779,
122
+ "model_unique_tensor_bytes": 67120,
123
+ "checkpoint_file_bytes": 73096,
124
+ "neural_parameters": 0,
125
+ "adaptation_seconds": 0.008381085004657507,
126
+ "replay_identical": true,
127
+ "all_adaptation_examples_source_only": true
128
+ },
129
+ {
130
+ "samples": 64,
131
+ "seed": 20851,
132
+ "model": "joint_gaussian",
133
+ "held_records": 404,
134
+ "complete_load_and_normalization_seconds": 0.0008900800021365285,
135
+ "inference_mean_seconds": 0.00026364240475231784,
136
+ "inference_p95_seconds": 0.00029619950219057496,
137
+ "model_unique_tensor_bytes": 39248,
138
+ "checkpoint_file_bytes": 43665,
139
+ "neural_parameters": 0,
140
+ "adaptation_seconds": 0.0005561420111916959,
141
+ "replay_identical": true,
142
+ "all_adaptation_examples_source_only": true
143
+ },
144
+ {
145
+ "samples": 64,
146
+ "seed": 20851,
147
+ "model": "factorized_logistic",
148
+ "held_records": 404,
149
+ "complete_load_and_normalization_seconds": 0.000848341005621478,
150
+ "inference_mean_seconds": 0.00012068490032106638,
151
+ "inference_p95_seconds": 0.00019450574545771802,
152
+ "model_unique_tensor_bytes": 3460,
153
+ "checkpoint_file_bytes": 7944,
154
+ "neural_parameters": 743,
155
+ "adaptation_seconds": 2.3290587439987576,
156
+ "replay_identical": true,
157
+ "all_adaptation_examples_source_only": true
158
+ },
159
+ {
160
+ "samples": 64,
161
+ "seed": 20851,
162
+ "model": "factorized_classifier",
163
+ "held_records": 404,
164
+ "complete_load_and_normalization_seconds": 0.0013938909978605807,
165
+ "inference_mean_seconds": 0.00014313119900180026,
166
+ "inference_p95_seconds": 0.0002015103484154678,
167
+ "model_unique_tensor_bytes": 80348,
168
+ "checkpoint_file_bytes": 85490,
169
+ "neural_parameters": 19965,
170
+ "adaptation_seconds": 2.0678669799963245,
171
+ "replay_identical": true,
172
+ "all_adaptation_examples_source_only": true
173
+ },
174
+ {
175
+ "samples": 64,
176
+ "seed": 20852,
177
+ "model": "additive_full",
178
+ "held_records": 404,
179
+ "complete_load_and_normalization_seconds": 0.000961917990935035,
180
+ "inference_mean_seconds": 0.0002628864007419907,
181
+ "inference_p95_seconds": 0.00029028019998804665,
182
+ "model_unique_tensor_bytes": 45520,
183
+ "checkpoint_file_bytes": 50120,
184
+ "neural_parameters": 0,
185
+ "adaptation_seconds": 0.0011382470111129805,
186
+ "replay_identical": true,
187
+ "all_adaptation_examples_source_only": true
188
+ },
189
+ {
190
+ "samples": 64,
191
+ "seed": 20852,
192
+ "model": "additive_diag",
193
+ "held_records": 404,
194
+ "complete_load_and_normalization_seconds": 0.0009563920029904693,
195
+ "inference_mean_seconds": 0.0002612571028294042,
196
+ "inference_p95_seconds": 0.00028605145416804584,
197
+ "model_unique_tensor_bytes": 45520,
198
+ "checkpoint_file_bytes": 50120,
199
+ "neural_parameters": 0,
200
+ "adaptation_seconds": 0.0008023369882721454,
201
+ "replay_identical": true,
202
+ "all_adaptation_examples_source_only": true
203
+ },
204
+ {
205
+ "samples": 64,
206
+ "seed": 20852,
207
+ "model": "factorized_neural",
208
+ "held_records": 404,
209
+ "complete_load_and_normalization_seconds": 0.0012270549923414364,
210
+ "inference_mean_seconds": 0.00026143819850403814,
211
+ "inference_p95_seconds": 0.0002890330950322095,
212
+ "model_unique_tensor_bytes": 118992,
213
+ "checkpoint_file_bytes": 124555,
214
+ "neural_parameters": 19936,
215
+ "adaptation_seconds": 1.4909278029954294,
216
+ "replay_identical": true,
217
+ "all_adaptation_examples_source_only": true
218
+ },
219
+ {
220
+ "samples": 64,
221
+ "seed": 20852,
222
+ "model": "joint_neural",
223
+ "held_records": 404,
224
+ "complete_load_and_normalization_seconds": 0.0012015550018986687,
225
+ "inference_mean_seconds": 0.00026175030070589854,
226
+ "inference_p95_seconds": 0.0002891753509175032,
227
+ "model_unique_tensor_bytes": 119164,
228
+ "checkpoint_file_bytes": 124672,
229
+ "neural_parameters": 19979,
230
+ "adaptation_seconds": 1.9436139250028646,
231
+ "replay_identical": true,
232
+ "all_adaptation_examples_source_only": true
233
+ },
234
+ {
235
+ "samples": 64,
236
+ "seed": 20852,
237
+ "model": "factorized_nb",
238
+ "held_records": 404,
239
+ "complete_load_and_normalization_seconds": 0.0013815489946864545,
240
+ "inference_mean_seconds": 0.0006097001998568885,
241
+ "inference_p95_seconds": 0.0006552168029884341,
242
+ "model_unique_tensor_bytes": 362248,
243
+ "checkpoint_file_bytes": 367954,
244
+ "neural_parameters": 0,
245
+ "adaptation_seconds": 0.002787183999316767,
246
+ "replay_identical": true,
247
+ "all_adaptation_examples_source_only": true
248
+ },
249
+ {
250
+ "samples": 64,
251
+ "seed": 20852,
252
+ "model": "factorized_lda",
253
+ "held_records": 404,
254
+ "complete_load_and_normalization_seconds": 0.0012272910098545253,
255
+ "inference_mean_seconds": 0.0006049915988114663,
256
+ "inference_p95_seconds": 0.0006384678956237621,
257
+ "model_unique_tensor_bytes": 86280,
258
+ "checkpoint_file_bytes": 92002,
259
+ "neural_parameters": 0,
260
+ "adaptation_seconds": 0.002600148000055924,
261
+ "replay_identical": true,
262
+ "all_adaptation_examples_source_only": true
263
+ },
264
+ {
265
+ "samples": 64,
266
+ "seed": 20852,
267
+ "model": "factorized_kernel",
268
+ "held_records": 404,
269
+ "complete_load_and_normalization_seconds": 0.0012309000012464821,
270
+ "inference_mean_seconds": 0.0026382880983874205,
271
+ "inference_p95_seconds": 0.002680203395721037,
272
+ "model_unique_tensor_bytes": 86280,
273
+ "checkpoint_file_bytes": 92050,
274
+ "neural_parameters": 0,
275
+ "adaptation_seconds": 0.0028671200125245377,
276
+ "replay_identical": true,
277
+ "all_adaptation_examples_source_only": true
278
+ },
279
+ {
280
+ "samples": 64,
281
+ "seed": 20852,
282
+ "model": "factorized_ridge",
283
+ "held_records": 404,
284
+ "complete_load_and_normalization_seconds": 0.0012845680030295625,
285
+ "inference_mean_seconds": 0.0002497603971278295,
286
+ "inference_p95_seconds": 0.00029410949719022023,
287
+ "model_unique_tensor_bytes": 67120,
288
+ "checkpoint_file_bytes": 73096,
289
+ "neural_parameters": 0,
290
+ "adaptation_seconds": 0.004647762005333789,
291
+ "replay_identical": true,
292
+ "all_adaptation_examples_source_only": true
293
+ },
294
+ {
295
+ "samples": 64,
296
+ "seed": 20852,
297
+ "model": "joint_gaussian",
298
+ "held_records": 404,
299
+ "complete_load_and_normalization_seconds": 0.0008687079971423373,
300
+ "inference_mean_seconds": 0.00025967070396291095,
301
+ "inference_p95_seconds": 0.000284310905408347,
302
+ "model_unique_tensor_bytes": 39248,
303
+ "checkpoint_file_bytes": 43665,
304
+ "neural_parameters": 0,
305
+ "adaptation_seconds": 0.001063668998540379,
306
+ "replay_identical": true,
307
+ "all_adaptation_examples_source_only": true
308
+ },
309
+ {
310
+ "samples": 64,
311
+ "seed": 20852,
312
+ "model": "factorized_logistic",
313
+ "held_records": 404,
314
+ "complete_load_and_normalization_seconds": 0.0008476429939037189,
315
+ "inference_mean_seconds": 0.00011152600054629148,
316
+ "inference_p95_seconds": 0.00013942135265097017,
317
+ "model_unique_tensor_bytes": 3460,
318
+ "checkpoint_file_bytes": 7944,
319
+ "neural_parameters": 743,
320
+ "adaptation_seconds": 2.3147749789932277,
321
+ "replay_identical": true,
322
+ "all_adaptation_examples_source_only": true
323
+ },
324
+ {
325
+ "samples": 64,
326
+ "seed": 20852,
327
+ "model": "factorized_classifier",
328
+ "held_records": 404,
329
+ "complete_load_and_normalization_seconds": 0.001020545998471789,
330
+ "inference_mean_seconds": 0.00013776839914498852,
331
+ "inference_p95_seconds": 0.00017045739732566285,
332
+ "model_unique_tensor_bytes": 80348,
333
+ "checkpoint_file_bytes": 85490,
334
+ "neural_parameters": 19965,
335
+ "adaptation_seconds": 2.5580791469983524,
336
+ "replay_identical": true,
337
+ "all_adaptation_examples_source_only": true
338
+ },
339
+ {
340
+ "samples": 64,
341
+ "seed": 20853,
342
+ "model": "additive_full",
343
+ "held_records": 404,
344
+ "complete_load_and_normalization_seconds": 0.0009024010068969801,
345
+ "inference_mean_seconds": 0.0002611572010209784,
346
+ "inference_p95_seconds": 0.000286553749901941,
347
+ "model_unique_tensor_bytes": 45520,
348
+ "checkpoint_file_bytes": 50120,
349
+ "neural_parameters": 0,
350
+ "adaptation_seconds": 0.0028005479980492964,
351
+ "replay_identical": true,
352
+ "all_adaptation_examples_source_only": true
353
+ },
354
+ {
355
+ "samples": 64,
356
+ "seed": 20853,
357
+ "model": "additive_diag",
358
+ "held_records": 404,
359
+ "complete_load_and_normalization_seconds": 0.0009209519921569154,
360
+ "inference_mean_seconds": 0.00026268590008839967,
361
+ "inference_p95_seconds": 0.0002892694479669444,
362
+ "model_unique_tensor_bytes": 45520,
363
+ "checkpoint_file_bytes": 50120,
364
+ "neural_parameters": 0,
365
+ "adaptation_seconds": 0.003586471008020453,
366
+ "replay_identical": true,
367
+ "all_adaptation_examples_source_only": true
368
+ },
369
+ {
370
+ "samples": 64,
371
+ "seed": 20853,
372
+ "model": "factorized_neural",
373
+ "held_records": 404,
374
+ "complete_load_and_normalization_seconds": 0.0011656420101644471,
375
+ "inference_mean_seconds": 0.00025993179879151284,
376
+ "inference_p95_seconds": 0.0002864694484742358,
377
+ "model_unique_tensor_bytes": 118992,
378
+ "checkpoint_file_bytes": 124555,
379
+ "neural_parameters": 19936,
380
+ "adaptation_seconds": 1.807837837986881,
381
+ "replay_identical": true,
382
+ "all_adaptation_examples_source_only": true
383
+ },
384
+ {
385
+ "samples": 64,
386
+ "seed": 20853,
387
+ "model": "joint_neural",
388
+ "held_records": 404,
389
+ "complete_load_and_normalization_seconds": 0.001180416002171114,
390
+ "inference_mean_seconds": 0.00026060349919134753,
391
+ "inference_p95_seconds": 0.0002875160396797582,
392
+ "model_unique_tensor_bytes": 119164,
393
+ "checkpoint_file_bytes": 124672,
394
+ "neural_parameters": 19979,
395
+ "adaptation_seconds": 1.819105454007513,
396
+ "replay_identical": true,
397
+ "all_adaptation_examples_source_only": true
398
+ },
399
+ {
400
+ "samples": 64,
401
+ "seed": 20853,
402
+ "model": "factorized_nb",
403
+ "held_records": 404,
404
+ "complete_load_and_normalization_seconds": 0.0013133949978509918,
405
+ "inference_mean_seconds": 0.0006065756970201619,
406
+ "inference_p95_seconds": 0.0006397515397111419,
407
+ "model_unique_tensor_bytes": 362248,
408
+ "checkpoint_file_bytes": 367954,
409
+ "neural_parameters": 0,
410
+ "adaptation_seconds": 0.0017201379960170016,
411
+ "replay_identical": true,
412
+ "all_adaptation_examples_source_only": true
413
+ },
414
+ {
415
+ "samples": 64,
416
+ "seed": 20853,
417
+ "model": "factorized_lda",
418
+ "held_records": 404,
419
+ "complete_load_and_normalization_seconds": 0.0012328849988989532,
420
+ "inference_mean_seconds": 0.0006044520981959068,
421
+ "inference_p95_seconds": 0.0006506119469122495,
422
+ "model_unique_tensor_bytes": 86280,
423
+ "checkpoint_file_bytes": 92002,
424
+ "neural_parameters": 0,
425
+ "adaptation_seconds": 0.001707875999272801,
426
+ "replay_identical": true,
427
+ "all_adaptation_examples_source_only": true
428
+ },
429
+ {
430
+ "samples": 64,
431
+ "seed": 20853,
432
+ "model": "factorized_kernel",
433
+ "held_records": 404,
434
+ "complete_load_and_normalization_seconds": 0.0012119750026613474,
435
+ "inference_mean_seconds": 0.002643710200209171,
436
+ "inference_p95_seconds": 0.00269921550352592,
437
+ "model_unique_tensor_bytes": 86280,
438
+ "checkpoint_file_bytes": 92050,
439
+ "neural_parameters": 0,
440
+ "adaptation_seconds": 0.0015314379998017102,
441
+ "replay_identical": true,
442
+ "all_adaptation_examples_source_only": true
443
+ },
444
+ {
445
+ "samples": 64,
446
+ "seed": 20853,
447
+ "model": "factorized_ridge",
448
+ "held_records": 404,
449
+ "complete_load_and_normalization_seconds": 0.001265758997760713,
450
+ "inference_mean_seconds": 0.00024933659879025074,
451
+ "inference_p95_seconds": 0.000285216694464907,
452
+ "model_unique_tensor_bytes": 67120,
453
+ "checkpoint_file_bytes": 73096,
454
+ "neural_parameters": 0,
455
+ "adaptation_seconds": 0.0024104670010274276,
456
+ "replay_identical": true,
457
+ "all_adaptation_examples_source_only": true
458
+ },
459
+ {
460
+ "samples": 64,
461
+ "seed": 20853,
462
+ "model": "joint_gaussian",
463
+ "held_records": 404,
464
+ "complete_load_and_normalization_seconds": 0.0008657400030642748,
465
+ "inference_mean_seconds": 0.00026037730276584625,
466
+ "inference_p95_seconds": 0.0002871448064979631,
467
+ "model_unique_tensor_bytes": 39248,
468
+ "checkpoint_file_bytes": 43665,
469
+ "neural_parameters": 0,
470
+ "adaptation_seconds": 0.0010624630085658282,
471
+ "replay_identical": true,
472
+ "all_adaptation_examples_source_only": true
473
+ },
474
+ {
475
+ "samples": 64,
476
+ "seed": 20853,
477
+ "model": "factorized_logistic",
478
+ "held_records": 404,
479
+ "complete_load_and_normalization_seconds": 0.0008373909950023517,
480
+ "inference_mean_seconds": 0.00011191850062459708,
481
+ "inference_p95_seconds": 0.0001402240006427746,
482
+ "model_unique_tensor_bytes": 3460,
483
+ "checkpoint_file_bytes": 7944,
484
+ "neural_parameters": 743,
485
+ "adaptation_seconds": 2.309631664000335,
486
+ "replay_identical": true,
487
+ "all_adaptation_examples_source_only": true
488
+ },
489
+ {
490
+ "samples": 64,
491
+ "seed": 20853,
492
+ "model": "factorized_classifier",
493
+ "held_records": 404,
494
+ "complete_load_and_normalization_seconds": 0.0010054969898192212,
495
+ "inference_mean_seconds": 0.00013763299793936312,
496
+ "inference_p95_seconds": 0.00017116849703597832,
497
+ "model_unique_tensor_bytes": 80348,
498
+ "checkpoint_file_bytes": 85490,
499
+ "neural_parameters": 19965,
500
+ "adaptation_seconds": 2.5695332850009436,
501
+ "replay_identical": true,
502
+ "all_adaptation_examples_source_only": true
503
+ },
504
+ {
505
+ "samples": 128,
506
+ "seed": 20851,
507
+ "model": "additive_full",
508
+ "held_records": 404,
509
+ "complete_load_and_normalization_seconds": 0.0009274089970858768,
510
+ "inference_mean_seconds": 0.00025984980311477557,
511
+ "inference_p95_seconds": 0.0002835436549503356,
512
+ "model_unique_tensor_bytes": 45520,
513
+ "checkpoint_file_bytes": 50632,
514
+ "neural_parameters": 0,
515
+ "adaptation_seconds": 0.0014137029938865453,
516
+ "replay_identical": true,
517
+ "all_adaptation_examples_source_only": true
518
+ },
519
+ {
520
+ "samples": 128,
521
+ "seed": 20851,
522
+ "model": "additive_diag",
523
+ "held_records": 404,
524
+ "complete_load_and_normalization_seconds": 0.0009071249951375648,
525
+ "inference_mean_seconds": 0.0002616522018797696,
526
+ "inference_p95_seconds": 0.000287238755117869,
527
+ "model_unique_tensor_bytes": 45520,
528
+ "checkpoint_file_bytes": 50632,
529
+ "neural_parameters": 0,
530
+ "adaptation_seconds": 0.0007983359973877668,
531
+ "replay_identical": true,
532
+ "all_adaptation_examples_source_only": true
533
+ },
534
+ {
535
+ "samples": 128,
536
+ "seed": 20851,
537
+ "model": "factorized_neural",
538
+ "held_records": 404,
539
+ "complete_load_and_normalization_seconds": 0.0011561970022739843,
540
+ "inference_mean_seconds": 0.0002601777989184484,
541
+ "inference_p95_seconds": 0.00028696715089608914,
542
+ "model_unique_tensor_bytes": 118992,
543
+ "checkpoint_file_bytes": 125067,
544
+ "neural_parameters": 19936,
545
+ "adaptation_seconds": 1.1472879990033107,
546
+ "replay_identical": true,
547
+ "all_adaptation_examples_source_only": true
548
+ },
549
+ {
550
+ "samples": 128,
551
+ "seed": 20851,
552
+ "model": "joint_neural",
553
+ "held_records": 404,
554
+ "complete_load_and_normalization_seconds": 0.0011630950029939413,
555
+ "inference_mean_seconds": 0.0002618298953166232,
556
+ "inference_p95_seconds": 0.00029468045104295013,
557
+ "model_unique_tensor_bytes": 119164,
558
+ "checkpoint_file_bytes": 125184,
559
+ "neural_parameters": 19979,
560
+ "adaptation_seconds": 1.1532195689942455,
561
+ "replay_identical": true,
562
+ "all_adaptation_examples_source_only": true
563
+ },
564
+ {
565
+ "samples": 128,
566
+ "seed": 20851,
567
+ "model": "factorized_nb",
568
+ "held_records": 404,
569
+ "complete_load_and_normalization_seconds": 0.0013315870019141585,
570
+ "inference_mean_seconds": 0.0005982018948998303,
571
+ "inference_p95_seconds": 0.000636294449941488,
572
+ "model_unique_tensor_bytes": 391944,
573
+ "checkpoint_file_bytes": 398162,
574
+ "neural_parameters": 0,
575
+ "adaptation_seconds": 0.0007647439924767241,
576
+ "replay_identical": true,
577
+ "all_adaptation_examples_source_only": true
578
+ },
579
+ {
580
+ "samples": 128,
581
+ "seed": 20851,
582
+ "model": "factorized_lda",
583
+ "held_records": 404,
584
+ "complete_load_and_normalization_seconds": 0.0012431110080797225,
585
+ "inference_mean_seconds": 0.000607154899626039,
586
+ "inference_p95_seconds": 0.0006498076952993869,
587
+ "model_unique_tensor_bytes": 115976,
588
+ "checkpoint_file_bytes": 122210,
589
+ "neural_parameters": 0,
590
+ "adaptation_seconds": 0.0009372979984618723,
591
+ "replay_identical": true,
592
+ "all_adaptation_examples_source_only": true
593
+ },
594
+ {
595
+ "samples": 128,
596
+ "seed": 20851,
597
+ "model": "factorized_kernel",
598
+ "held_records": 404,
599
+ "complete_load_and_normalization_seconds": 0.001224384002853185,
600
+ "inference_mean_seconds": 0.0027609770026174375,
601
+ "inference_p95_seconds": 0.002805410551809473,
602
+ "model_unique_tensor_bytes": 115976,
603
+ "checkpoint_file_bytes": 122258,
604
+ "neural_parameters": 0,
605
+ "adaptation_seconds": 0.0008651509997434914,
606
+ "replay_identical": true,
607
+ "all_adaptation_examples_source_only": true
608
+ },
609
+ {
610
+ "samples": 128,
611
+ "seed": 20851,
612
+ "model": "factorized_ridge",
613
+ "held_records": 404,
614
+ "complete_load_and_normalization_seconds": 0.0012712500029010698,
615
+ "inference_mean_seconds": 0.00024874110094970093,
616
+ "inference_p95_seconds": 0.00028713019637507383,
617
+ "model_unique_tensor_bytes": 96816,
618
+ "checkpoint_file_bytes": 103304,
619
+ "neural_parameters": 0,
620
+ "adaptation_seconds": 0.001426750997779891,
621
+ "replay_identical": true,
622
+ "all_adaptation_examples_source_only": true
623
+ },
624
+ {
625
+ "samples": 128,
626
+ "seed": 20851,
627
+ "model": "joint_gaussian",
628
+ "held_records": 404,
629
+ "complete_load_and_normalization_seconds": 0.0008823450043564662,
630
+ "inference_mean_seconds": 0.00026123360003111885,
631
+ "inference_p95_seconds": 0.0002926409535575658,
632
+ "model_unique_tensor_bytes": 39248,
633
+ "checkpoint_file_bytes": 44177,
634
+ "neural_parameters": 0,
635
+ "adaptation_seconds": 0.0005419520020950586,
636
+ "replay_identical": true,
637
+ "all_adaptation_examples_source_only": true
638
+ },
639
+ {
640
+ "samples": 128,
641
+ "seed": 20851,
642
+ "model": "factorized_logistic",
643
+ "held_records": 404,
644
+ "complete_load_and_normalization_seconds": 0.0008323160000145435,
645
+ "inference_mean_seconds": 0.00011202480091014877,
646
+ "inference_p95_seconds": 0.00014472635448328213,
647
+ "model_unique_tensor_bytes": 3460,
648
+ "checkpoint_file_bytes": 8456,
649
+ "neural_parameters": 743,
650
+ "adaptation_seconds": 2.265729534003185,
651
+ "replay_identical": true,
652
+ "all_adaptation_examples_source_only": true
653
+ },
654
+ {
655
+ "samples": 128,
656
+ "seed": 20851,
657
+ "model": "factorized_classifier",
658
+ "held_records": 404,
659
+ "complete_load_and_normalization_seconds": 0.0010060169879579917,
660
+ "inference_mean_seconds": 0.0001379326029564254,
661
+ "inference_p95_seconds": 0.00016989460200420576,
662
+ "model_unique_tensor_bytes": 80348,
663
+ "checkpoint_file_bytes": 86002,
664
+ "neural_parameters": 19965,
665
+ "adaptation_seconds": 2.539189789997181,
666
+ "replay_identical": true,
667
+ "all_adaptation_examples_source_only": true
668
+ },
669
+ {
670
+ "samples": 128,
671
+ "seed": 20852,
672
+ "model": "additive_full",
673
+ "held_records": 404,
674
+ "complete_load_and_normalization_seconds": 0.000928383000427857,
675
+ "inference_mean_seconds": 0.000262654697871767,
676
+ "inference_p95_seconds": 0.0002870331998565234,
677
+ "model_unique_tensor_bytes": 45520,
678
+ "checkpoint_file_bytes": 50632,
679
+ "neural_parameters": 0,
680
+ "adaptation_seconds": 0.0011084300058428198,
681
+ "replay_identical": true,
682
+ "all_adaptation_examples_source_only": true
683
+ },
684
+ {
685
+ "samples": 128,
686
+ "seed": 20852,
687
+ "model": "additive_diag",
688
+ "held_records": 404,
689
+ "complete_load_and_normalization_seconds": 0.0009170060075121,
690
+ "inference_mean_seconds": 0.00026034649781649934,
691
+ "inference_p95_seconds": 0.0002838792985130567,
692
+ "model_unique_tensor_bytes": 45520,
693
+ "checkpoint_file_bytes": 50632,
694
+ "neural_parameters": 0,
695
+ "adaptation_seconds": 0.0008065050060395151,
696
+ "replay_identical": true,
697
+ "all_adaptation_examples_source_only": true
698
+ },
699
+ {
700
+ "samples": 128,
701
+ "seed": 20852,
702
+ "model": "factorized_neural",
703
+ "held_records": 404,
704
+ "complete_load_and_normalization_seconds": 0.0011729620018741116,
705
+ "inference_mean_seconds": 0.0002620630024466664,
706
+ "inference_p95_seconds": 0.00029578709581983275,
707
+ "model_unique_tensor_bytes": 118992,
708
+ "checkpoint_file_bytes": 125067,
709
+ "neural_parameters": 19936,
710
+ "adaptation_seconds": 1.7587256989936577,
711
+ "replay_identical": true,
712
+ "all_adaptation_examples_source_only": true
713
+ },
714
+ {
715
+ "samples": 128,
716
+ "seed": 20852,
717
+ "model": "joint_neural",
718
+ "held_records": 404,
719
+ "complete_load_and_normalization_seconds": 0.0011583169980440289,
720
+ "inference_mean_seconds": 0.0002600443971459754,
721
+ "inference_p95_seconds": 0.00029063429828966034,
722
+ "model_unique_tensor_bytes": 119164,
723
+ "checkpoint_file_bytes": 125184,
724
+ "neural_parameters": 19979,
725
+ "adaptation_seconds": 1.8775172630121233,
726
+ "replay_identical": true,
727
+ "all_adaptation_examples_source_only": true
728
+ },
729
+ {
730
+ "samples": 128,
731
+ "seed": 20852,
732
+ "model": "factorized_nb",
733
+ "held_records": 404,
734
+ "complete_load_and_normalization_seconds": 0.00132534300792031,
735
+ "inference_mean_seconds": 0.0006039863015757874,
736
+ "inference_p95_seconds": 0.0006463250530941877,
737
+ "model_unique_tensor_bytes": 391944,
738
+ "checkpoint_file_bytes": 398162,
739
+ "neural_parameters": 0,
740
+ "adaptation_seconds": 0.0007821409963071346,
741
+ "replay_identical": true,
742
+ "all_adaptation_examples_source_only": true
743
+ },
744
+ {
745
+ "samples": 128,
746
+ "seed": 20852,
747
+ "model": "factorized_lda",
748
+ "held_records": 404,
749
+ "complete_load_and_normalization_seconds": 0.0012198199983686209,
750
+ "inference_mean_seconds": 0.0006062720000045374,
751
+ "inference_p95_seconds": 0.0006462112985900603,
752
+ "model_unique_tensor_bytes": 115976,
753
+ "checkpoint_file_bytes": 122210,
754
+ "neural_parameters": 0,
755
+ "adaptation_seconds": 0.0006677630008198321,
756
+ "replay_identical": true,
757
+ "all_adaptation_examples_source_only": true
758
+ },
759
+ {
760
+ "samples": 128,
761
+ "seed": 20852,
762
+ "model": "factorized_kernel",
763
+ "held_records": 404,
764
+ "complete_load_and_normalization_seconds": 0.0012083490000804886,
765
+ "inference_mean_seconds": 0.0027812719985377044,
766
+ "inference_p95_seconds": 0.002826895395264728,
767
+ "model_unique_tensor_bytes": 115976,
768
+ "checkpoint_file_bytes": 122258,
769
+ "neural_parameters": 0,
770
+ "adaptation_seconds": 0.0007488839910365641,
771
+ "replay_identical": true,
772
+ "all_adaptation_examples_source_only": true
773
+ },
774
+ {
775
+ "samples": 128,
776
+ "seed": 20852,
777
+ "model": "factorized_ridge",
778
+ "held_records": 404,
779
+ "complete_load_and_normalization_seconds": 0.0012828539911424741,
780
+ "inference_mean_seconds": 0.0002481356976204552,
781
+ "inference_p95_seconds": 0.0002843516980647109,
782
+ "model_unique_tensor_bytes": 96816,
783
+ "checkpoint_file_bytes": 103304,
784
+ "neural_parameters": 0,
785
+ "adaptation_seconds": 0.0011679009912768379,
786
+ "replay_identical": true,
787
+ "all_adaptation_examples_source_only": true
788
+ },
789
+ {
790
+ "samples": 128,
791
+ "seed": 20852,
792
+ "model": "joint_gaussian",
793
+ "held_records": 404,
794
+ "complete_load_and_normalization_seconds": 0.0008719130128156394,
795
+ "inference_mean_seconds": 0.0002624836983159185,
796
+ "inference_p95_seconds": 0.00029371084528975185,
797
+ "model_unique_tensor_bytes": 39248,
798
+ "checkpoint_file_bytes": 44177,
799
+ "neural_parameters": 0,
800
+ "adaptation_seconds": 0.0005162120069144294,
801
+ "replay_identical": true,
802
+ "all_adaptation_examples_source_only": true
803
+ },
804
+ {
805
+ "samples": 128,
806
+ "seed": 20852,
807
+ "model": "factorized_logistic",
808
+ "held_records": 404,
809
+ "complete_load_and_normalization_seconds": 0.0008248350059147924,
810
+ "inference_mean_seconds": 0.00011078920069849119,
811
+ "inference_p95_seconds": 0.0001388419543218333,
812
+ "model_unique_tensor_bytes": 3460,
813
+ "checkpoint_file_bytes": 8456,
814
+ "neural_parameters": 743,
815
+ "adaptation_seconds": 2.2780268480128143,
816
+ "replay_identical": true,
817
+ "all_adaptation_examples_source_only": true
818
+ },
819
+ {
820
+ "samples": 128,
821
+ "seed": 20852,
822
+ "model": "factorized_classifier",
823
+ "held_records": 404,
824
+ "complete_load_and_normalization_seconds": 0.0010127889981959015,
825
+ "inference_mean_seconds": 0.0001387602009344846,
826
+ "inference_p95_seconds": 0.00017046920038410457,
827
+ "model_unique_tensor_bytes": 80348,
828
+ "checkpoint_file_bytes": 86002,
829
+ "neural_parameters": 19965,
830
+ "adaptation_seconds": 2.5505702320078854,
831
+ "replay_identical": true,
832
+ "all_adaptation_examples_source_only": true
833
+ },
834
+ {
835
+ "samples": 128,
836
+ "seed": 20853,
837
+ "model": "additive_full",
838
+ "held_records": 404,
839
+ "complete_load_and_normalization_seconds": 0.0008985570020740852,
840
+ "inference_mean_seconds": 0.0002609356000903063,
841
+ "inference_p95_seconds": 0.00028576055847224775,
842
+ "model_unique_tensor_bytes": 45520,
843
+ "checkpoint_file_bytes": 50632,
844
+ "neural_parameters": 0,
845
+ "adaptation_seconds": 0.0011253239936195314,
846
+ "replay_identical": true,
847
+ "all_adaptation_examples_source_only": true
848
+ },
849
+ {
850
+ "samples": 128,
851
+ "seed": 20853,
852
+ "model": "additive_diag",
853
+ "held_records": 404,
854
+ "complete_load_and_normalization_seconds": 0.0008946640009526163,
855
+ "inference_mean_seconds": 0.00026203940069535747,
856
+ "inference_p95_seconds": 0.00028610795197892,
857
+ "model_unique_tensor_bytes": 45520,
858
+ "checkpoint_file_bytes": 50632,
859
+ "neural_parameters": 0,
860
+ "adaptation_seconds": 0.0008097370009636506,
861
+ "replay_identical": true,
862
+ "all_adaptation_examples_source_only": true
863
+ },
864
+ {
865
+ "samples": 128,
866
+ "seed": 20853,
867
+ "model": "factorized_neural",
868
+ "held_records": 404,
869
+ "complete_load_and_normalization_seconds": 0.0011427699937485158,
870
+ "inference_mean_seconds": 0.0002616086014313623,
871
+ "inference_p95_seconds": 0.00029375480298767794,
872
+ "model_unique_tensor_bytes": 118992,
873
+ "checkpoint_file_bytes": 125067,
874
+ "neural_parameters": 19936,
875
+ "adaptation_seconds": 1.9229610209877137,
876
+ "replay_identical": true,
877
+ "all_adaptation_examples_source_only": true
878
+ },
879
+ {
880
+ "samples": 128,
881
+ "seed": 20853,
882
+ "model": "joint_neural",
883
+ "held_records": 404,
884
+ "complete_load_and_normalization_seconds": 0.001167598005849868,
885
+ "inference_mean_seconds": 0.0002601696993224323,
886
+ "inference_p95_seconds": 0.00028581810256582684,
887
+ "model_unique_tensor_bytes": 119164,
888
+ "checkpoint_file_bytes": 125184,
889
+ "neural_parameters": 19979,
890
+ "adaptation_seconds": 1.8031580100068823,
891
+ "replay_identical": true,
892
+ "all_adaptation_examples_source_only": true
893
+ },
894
+ {
895
+ "samples": 128,
896
+ "seed": 20853,
897
+ "model": "factorized_nb",
898
+ "held_records": 404,
899
+ "complete_load_and_normalization_seconds": 0.0013061780045973137,
900
+ "inference_mean_seconds": 0.0006047445014701225,
901
+ "inference_p95_seconds": 0.0006453416557633318,
902
+ "model_unique_tensor_bytes": 391944,
903
+ "checkpoint_file_bytes": 398162,
904
+ "neural_parameters": 0,
905
+ "adaptation_seconds": 0.0007809509988874197,
906
+ "replay_identical": true,
907
+ "all_adaptation_examples_source_only": true
908
+ },
909
+ {
910
+ "samples": 128,
911
+ "seed": 20853,
912
+ "model": "factorized_lda",
913
+ "held_records": 404,
914
+ "complete_load_and_normalization_seconds": 0.0012173220020486042,
915
+ "inference_mean_seconds": 0.000605196098331362,
916
+ "inference_p95_seconds": 0.0006497378461062908,
917
+ "model_unique_tensor_bytes": 115976,
918
+ "checkpoint_file_bytes": 122210,
919
+ "neural_parameters": 0,
920
+ "adaptation_seconds": 0.0006537870067404583,
921
+ "replay_identical": true,
922
+ "all_adaptation_examples_source_only": true
923
+ },
924
+ {
925
+ "samples": 128,
926
+ "seed": 20853,
927
+ "model": "factorized_kernel",
928
+ "held_records": 404,
929
+ "complete_load_and_normalization_seconds": 0.001217095006722957,
930
+ "inference_mean_seconds": 0.002776125697710086,
931
+ "inference_p95_seconds": 0.002839853251498425,
932
+ "model_unique_tensor_bytes": 115976,
933
+ "checkpoint_file_bytes": 122258,
934
+ "neural_parameters": 0,
935
+ "adaptation_seconds": 0.0007469249976566061,
936
+ "replay_identical": true,
937
+ "all_adaptation_examples_source_only": true
938
+ },
939
+ {
940
+ "samples": 128,
941
+ "seed": 20853,
942
+ "model": "factorized_ridge",
943
+ "held_records": 404,
944
+ "complete_load_and_normalization_seconds": 0.0012772000045515597,
945
+ "inference_mean_seconds": 0.0002497403998859227,
946
+ "inference_p95_seconds": 0.00028863170300610355,
947
+ "model_unique_tensor_bytes": 96816,
948
+ "checkpoint_file_bytes": 103304,
949
+ "neural_parameters": 0,
950
+ "adaptation_seconds": 0.001129559997934848,
951
+ "replay_identical": true,
952
+ "all_adaptation_examples_source_only": true
953
+ },
954
+ {
955
+ "samples": 128,
956
+ "seed": 20853,
957
+ "model": "joint_gaussian",
958
+ "held_records": 404,
959
+ "complete_load_and_normalization_seconds": 0.0008757549949223176,
960
+ "inference_mean_seconds": 0.0002603872009785846,
961
+ "inference_p95_seconds": 0.0002872442026273347,
962
+ "model_unique_tensor_bytes": 39248,
963
+ "checkpoint_file_bytes": 44177,
964
+ "neural_parameters": 0,
965
+ "adaptation_seconds": 0.0005088479956611991,
966
+ "replay_identical": true,
967
+ "all_adaptation_examples_source_only": true
968
+ },
969
+ {
970
+ "samples": 128,
971
+ "seed": 20853,
972
+ "model": "factorized_logistic",
973
+ "held_records": 404,
974
+ "complete_load_and_normalization_seconds": 0.0008460310054942966,
975
+ "inference_mean_seconds": 0.00011123630101792515,
976
+ "inference_p95_seconds": 0.00014035790663911025,
977
+ "model_unique_tensor_bytes": 3460,
978
+ "checkpoint_file_bytes": 8456,
979
+ "neural_parameters": 743,
980
+ "adaptation_seconds": 2.3297772530058865,
981
+ "replay_identical": true,
982
+ "all_adaptation_examples_source_only": true
983
+ },
984
+ {
985
+ "samples": 128,
986
+ "seed": 20853,
987
+ "model": "factorized_classifier",
988
+ "held_records": 404,
989
+ "complete_load_and_normalization_seconds": 0.0010084430105052888,
990
+ "inference_mean_seconds": 0.0001375074964016676,
991
+ "inference_p95_seconds": 0.00016899065085453907,
992
+ "model_unique_tensor_bytes": 80348,
993
+ "checkpoint_file_bytes": 86002,
994
+ "neural_parameters": 19965,
995
+ "adaptation_seconds": 2.581558428995777,
996
+ "replay_identical": true,
997
+ "all_adaptation_examples_source_only": true
998
+ }
999
+ ],
1000
+ "replay_failures": [],
1001
+ "fs_cache_warm": true,
1002
+ "no_new_fit_or_selection": true,
1003
+ "full_batch_latency_not_single_incident_latency": true,
1004
+ "original_stage_timers_exclude_process_startup": true,
1005
+ "physical_energy_and_transport_not_measured": true
1006
+ }
provenance/temperature/forecast_audit.json ADDED
@@ -0,0 +1,1809 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "rows": [
3
+ {
4
+ "samples": 64,
5
+ "seed": 20851,
6
+ "model": "additive_full",
7
+ "stratum": "unseen",
8
+ "observed_coordinates": 0,
9
+ "remaining_coordinates": 56,
10
+ "records": 65,
11
+ "forecast_nll_per_remaining_dimension": 0.5770275619754338
12
+ },
13
+ {
14
+ "samples": 64,
15
+ "seed": 20851,
16
+ "model": "additive_full",
17
+ "stratum": "seen",
18
+ "observed_coordinates": 0,
19
+ "remaining_coordinates": 56,
20
+ "records": 339,
21
+ "forecast_nll_per_remaining_dimension": 0.5737865699496951
22
+ },
23
+ {
24
+ "samples": 64,
25
+ "seed": 20851,
26
+ "model": "additive_full",
27
+ "stratum": "unseen",
28
+ "observed_coordinates": 8,
29
+ "remaining_coordinates": 48,
30
+ "records": 65,
31
+ "forecast_nll_per_remaining_dimension": 0.5256866205392998
32
+ },
33
+ {
34
+ "samples": 64,
35
+ "seed": 20851,
36
+ "model": "additive_full",
37
+ "stratum": "seen",
38
+ "observed_coordinates": 8,
39
+ "remaining_coordinates": 48,
40
+ "records": 339,
41
+ "forecast_nll_per_remaining_dimension": 0.5316446443489349
42
+ },
43
+ {
44
+ "samples": 64,
45
+ "seed": 20851,
46
+ "model": "additive_full",
47
+ "stratum": "unseen",
48
+ "observed_coordinates": 28,
49
+ "remaining_coordinates": 28,
50
+ "records": 65,
51
+ "forecast_nll_per_remaining_dimension": 0.43684891218151867
52
+ },
53
+ {
54
+ "samples": 64,
55
+ "seed": 20851,
56
+ "model": "additive_full",
57
+ "stratum": "seen",
58
+ "observed_coordinates": 28,
59
+ "remaining_coordinates": 28,
60
+ "records": 339,
61
+ "forecast_nll_per_remaining_dimension": 0.43475042426841554
62
+ },
63
+ {
64
+ "samples": 64,
65
+ "seed": 20851,
66
+ "model": "additive_diag",
67
+ "stratum": "unseen",
68
+ "observed_coordinates": 0,
69
+ "remaining_coordinates": 56,
70
+ "records": 65,
71
+ "forecast_nll_per_remaining_dimension": 1.0621006074507817
72
+ },
73
+ {
74
+ "samples": 64,
75
+ "seed": 20851,
76
+ "model": "additive_diag",
77
+ "stratum": "seen",
78
+ "observed_coordinates": 0,
79
+ "remaining_coordinates": 56,
80
+ "records": 339,
81
+ "forecast_nll_per_remaining_dimension": 1.031399672082987
82
+ },
83
+ {
84
+ "samples": 64,
85
+ "seed": 20851,
86
+ "model": "additive_diag",
87
+ "stratum": "unseen",
88
+ "observed_coordinates": 8,
89
+ "remaining_coordinates": 48,
90
+ "records": 65,
91
+ "forecast_nll_per_remaining_dimension": 1.019279419093396
92
+ },
93
+ {
94
+ "samples": 64,
95
+ "seed": 20851,
96
+ "model": "additive_diag",
97
+ "stratum": "seen",
98
+ "observed_coordinates": 8,
99
+ "remaining_coordinates": 48,
100
+ "records": 339,
101
+ "forecast_nll_per_remaining_dimension": 1.010876520043395
102
+ },
103
+ {
104
+ "samples": 64,
105
+ "seed": 20851,
106
+ "model": "additive_diag",
107
+ "stratum": "unseen",
108
+ "observed_coordinates": 28,
109
+ "remaining_coordinates": 28,
110
+ "records": 65,
111
+ "forecast_nll_per_remaining_dimension": 0.9329126433553542
112
+ },
113
+ {
114
+ "samples": 64,
115
+ "seed": 20851,
116
+ "model": "additive_diag",
117
+ "stratum": "seen",
118
+ "observed_coordinates": 28,
119
+ "remaining_coordinates": 28,
120
+ "records": 339,
121
+ "forecast_nll_per_remaining_dimension": 0.9539584217688821
122
+ },
123
+ {
124
+ "samples": 64,
125
+ "seed": 20851,
126
+ "model": "factorized_neural",
127
+ "stratum": "unseen",
128
+ "observed_coordinates": 0,
129
+ "remaining_coordinates": 56,
130
+ "records": 65,
131
+ "forecast_nll_per_remaining_dimension": 0.3545743945982203
132
+ },
133
+ {
134
+ "samples": 64,
135
+ "seed": 20851,
136
+ "model": "factorized_neural",
137
+ "stratum": "seen",
138
+ "observed_coordinates": 0,
139
+ "remaining_coordinates": 56,
140
+ "records": 339,
141
+ "forecast_nll_per_remaining_dimension": 0.3424401573542078
142
+ },
143
+ {
144
+ "samples": 64,
145
+ "seed": 20851,
146
+ "model": "factorized_neural",
147
+ "stratum": "unseen",
148
+ "observed_coordinates": 8,
149
+ "remaining_coordinates": 48,
150
+ "records": 65,
151
+ "forecast_nll_per_remaining_dimension": 0.28290233887383825
152
+ },
153
+ {
154
+ "samples": 64,
155
+ "seed": 20851,
156
+ "model": "factorized_neural",
157
+ "stratum": "seen",
158
+ "observed_coordinates": 8,
159
+ "remaining_coordinates": 48,
160
+ "records": 339,
161
+ "forecast_nll_per_remaining_dimension": 0.2728157586091853
162
+ },
163
+ {
164
+ "samples": 64,
165
+ "seed": 20851,
166
+ "model": "factorized_neural",
167
+ "stratum": "unseen",
168
+ "observed_coordinates": 28,
169
+ "remaining_coordinates": 28,
170
+ "records": 65,
171
+ "forecast_nll_per_remaining_dimension": 0.0484470454058143
172
+ },
173
+ {
174
+ "samples": 64,
175
+ "seed": 20851,
176
+ "model": "factorized_neural",
177
+ "stratum": "seen",
178
+ "observed_coordinates": 28,
179
+ "remaining_coordinates": 28,
180
+ "records": 339,
181
+ "forecast_nll_per_remaining_dimension": 0.02701689192152321
182
+ },
183
+ {
184
+ "samples": 64,
185
+ "seed": 20851,
186
+ "model": "joint_neural",
187
+ "stratum": "unseen",
188
+ "observed_coordinates": 0,
189
+ "remaining_coordinates": 56,
190
+ "records": 65,
191
+ "forecast_nll_per_remaining_dimension": 0.45355939918584603
192
+ },
193
+ {
194
+ "samples": 64,
195
+ "seed": 20851,
196
+ "model": "joint_neural",
197
+ "stratum": "seen",
198
+ "observed_coordinates": 0,
199
+ "remaining_coordinates": 56,
200
+ "records": 339,
201
+ "forecast_nll_per_remaining_dimension": 0.44716730143390476
202
+ },
203
+ {
204
+ "samples": 64,
205
+ "seed": 20851,
206
+ "model": "joint_neural",
207
+ "stratum": "unseen",
208
+ "observed_coordinates": 8,
209
+ "remaining_coordinates": 48,
210
+ "records": 65,
211
+ "forecast_nll_per_remaining_dimension": 0.39195012264945406
212
+ },
213
+ {
214
+ "samples": 64,
215
+ "seed": 20851,
216
+ "model": "joint_neural",
217
+ "stratum": "seen",
218
+ "observed_coordinates": 8,
219
+ "remaining_coordinates": 48,
220
+ "records": 339,
221
+ "forecast_nll_per_remaining_dimension": 0.3907787760683323
222
+ },
223
+ {
224
+ "samples": 64,
225
+ "seed": 20851,
226
+ "model": "joint_neural",
227
+ "stratum": "unseen",
228
+ "observed_coordinates": 28,
229
+ "remaining_coordinates": 28,
230
+ "records": 65,
231
+ "forecast_nll_per_remaining_dimension": 0.2187157033019343
232
+ },
233
+ {
234
+ "samples": 64,
235
+ "seed": 20851,
236
+ "model": "joint_neural",
237
+ "stratum": "seen",
238
+ "observed_coordinates": 28,
239
+ "remaining_coordinates": 28,
240
+ "records": 339,
241
+ "forecast_nll_per_remaining_dimension": 0.2092453671656058
242
+ },
243
+ {
244
+ "samples": 64,
245
+ "seed": 20851,
246
+ "model": "joint_gaussian",
247
+ "stratum": "unseen",
248
+ "observed_coordinates": 0,
249
+ "remaining_coordinates": 56,
250
+ "records": 65,
251
+ "forecast_nll_per_remaining_dimension": 0.49285368686449765
252
+ },
253
+ {
254
+ "samples": 64,
255
+ "seed": 20851,
256
+ "model": "joint_gaussian",
257
+ "stratum": "seen",
258
+ "observed_coordinates": 0,
259
+ "remaining_coordinates": 56,
260
+ "records": 339,
261
+ "forecast_nll_per_remaining_dimension": 0.49182121570933857
262
+ },
263
+ {
264
+ "samples": 64,
265
+ "seed": 20851,
266
+ "model": "joint_gaussian",
267
+ "stratum": "unseen",
268
+ "observed_coordinates": 8,
269
+ "remaining_coordinates": 48,
270
+ "records": 65,
271
+ "forecast_nll_per_remaining_dimension": 0.4344797976170246
272
+ },
273
+ {
274
+ "samples": 64,
275
+ "seed": 20851,
276
+ "model": "joint_gaussian",
277
+ "stratum": "seen",
278
+ "observed_coordinates": 8,
279
+ "remaining_coordinates": 48,
280
+ "records": 339,
281
+ "forecast_nll_per_remaining_dimension": 0.442010579808509
282
+ },
283
+ {
284
+ "samples": 64,
285
+ "seed": 20851,
286
+ "model": "joint_gaussian",
287
+ "stratum": "unseen",
288
+ "observed_coordinates": 28,
289
+ "remaining_coordinates": 28,
290
+ "records": 65,
291
+ "forecast_nll_per_remaining_dimension": 0.30237489372977405
292
+ },
293
+ {
294
+ "samples": 64,
295
+ "seed": 20851,
296
+ "model": "joint_gaussian",
297
+ "stratum": "seen",
298
+ "observed_coordinates": 28,
299
+ "remaining_coordinates": 28,
300
+ "records": 339,
301
+ "forecast_nll_per_remaining_dimension": 0.3002688051730416
302
+ },
303
+ {
304
+ "samples": 64,
305
+ "seed": 20852,
306
+ "model": "additive_full",
307
+ "stratum": "unseen",
308
+ "observed_coordinates": 0,
309
+ "remaining_coordinates": 56,
310
+ "records": 65,
311
+ "forecast_nll_per_remaining_dimension": 0.5580393556174009
312
+ },
313
+ {
314
+ "samples": 64,
315
+ "seed": 20852,
316
+ "model": "additive_full",
317
+ "stratum": "seen",
318
+ "observed_coordinates": 0,
319
+ "remaining_coordinates": 56,
320
+ "records": 339,
321
+ "forecast_nll_per_remaining_dimension": 0.5541405098546663
322
+ },
323
+ {
324
+ "samples": 64,
325
+ "seed": 20852,
326
+ "model": "additive_full",
327
+ "stratum": "unseen",
328
+ "observed_coordinates": 8,
329
+ "remaining_coordinates": 48,
330
+ "records": 65,
331
+ "forecast_nll_per_remaining_dimension": 0.500395015258323
332
+ },
333
+ {
334
+ "samples": 64,
335
+ "seed": 20852,
336
+ "model": "additive_full",
337
+ "stratum": "seen",
338
+ "observed_coordinates": 8,
339
+ "remaining_coordinates": 48,
340
+ "records": 339,
341
+ "forecast_nll_per_remaining_dimension": 0.5097716403635367
342
+ },
343
+ {
344
+ "samples": 64,
345
+ "seed": 20852,
346
+ "model": "additive_full",
347
+ "stratum": "unseen",
348
+ "observed_coordinates": 28,
349
+ "remaining_coordinates": 28,
350
+ "records": 65,
351
+ "forecast_nll_per_remaining_dimension": 0.4087553307845014
352
+ },
353
+ {
354
+ "samples": 64,
355
+ "seed": 20852,
356
+ "model": "additive_full",
357
+ "stratum": "seen",
358
+ "observed_coordinates": 28,
359
+ "remaining_coordinates": 28,
360
+ "records": 339,
361
+ "forecast_nll_per_remaining_dimension": 0.4119361718409499
362
+ },
363
+ {
364
+ "samples": 64,
365
+ "seed": 20852,
366
+ "model": "additive_diag",
367
+ "stratum": "unseen",
368
+ "observed_coordinates": 0,
369
+ "remaining_coordinates": 56,
370
+ "records": 65,
371
+ "forecast_nll_per_remaining_dimension": 1.0833457948235383
372
+ },
373
+ {
374
+ "samples": 64,
375
+ "seed": 20852,
376
+ "model": "additive_diag",
377
+ "stratum": "seen",
378
+ "observed_coordinates": 0,
379
+ "remaining_coordinates": 56,
380
+ "records": 339,
381
+ "forecast_nll_per_remaining_dimension": 1.0481185838324887
382
+ },
383
+ {
384
+ "samples": 64,
385
+ "seed": 20852,
386
+ "model": "additive_diag",
387
+ "stratum": "unseen",
388
+ "observed_coordinates": 8,
389
+ "remaining_coordinates": 48,
390
+ "records": 65,
391
+ "forecast_nll_per_remaining_dimension": 1.025779544476179
392
+ },
393
+ {
394
+ "samples": 64,
395
+ "seed": 20852,
396
+ "model": "additive_diag",
397
+ "stratum": "seen",
398
+ "observed_coordinates": 8,
399
+ "remaining_coordinates": 48,
400
+ "records": 339,
401
+ "forecast_nll_per_remaining_dimension": 1.0282678904166629
402
+ },
403
+ {
404
+ "samples": 64,
405
+ "seed": 20852,
406
+ "model": "additive_diag",
407
+ "stratum": "unseen",
408
+ "observed_coordinates": 28,
409
+ "remaining_coordinates": 28,
410
+ "records": 65,
411
+ "forecast_nll_per_remaining_dimension": 0.8293596806362908
412
+ },
413
+ {
414
+ "samples": 64,
415
+ "seed": 20852,
416
+ "model": "additive_diag",
417
+ "stratum": "seen",
418
+ "observed_coordinates": 28,
419
+ "remaining_coordinates": 28,
420
+ "records": 339,
421
+ "forecast_nll_per_remaining_dimension": 0.9284972710923115
422
+ },
423
+ {
424
+ "samples": 64,
425
+ "seed": 20852,
426
+ "model": "factorized_neural",
427
+ "stratum": "unseen",
428
+ "observed_coordinates": 0,
429
+ "remaining_coordinates": 56,
430
+ "records": 65,
431
+ "forecast_nll_per_remaining_dimension": 0.49922766108315497
432
+ },
433
+ {
434
+ "samples": 64,
435
+ "seed": 20852,
436
+ "model": "factorized_neural",
437
+ "stratum": "seen",
438
+ "observed_coordinates": 0,
439
+ "remaining_coordinates": 56,
440
+ "records": 339,
441
+ "forecast_nll_per_remaining_dimension": 0.4869317131351503
442
+ },
443
+ {
444
+ "samples": 64,
445
+ "seed": 20852,
446
+ "model": "factorized_neural",
447
+ "stratum": "unseen",
448
+ "observed_coordinates": 8,
449
+ "remaining_coordinates": 48,
450
+ "records": 65,
451
+ "forecast_nll_per_remaining_dimension": 0.42627195644832516
452
+ },
453
+ {
454
+ "samples": 64,
455
+ "seed": 20852,
456
+ "model": "factorized_neural",
457
+ "stratum": "seen",
458
+ "observed_coordinates": 8,
459
+ "remaining_coordinates": 48,
460
+ "records": 339,
461
+ "forecast_nll_per_remaining_dimension": 0.41854480398061544
462
+ },
463
+ {
464
+ "samples": 64,
465
+ "seed": 20852,
466
+ "model": "factorized_neural",
467
+ "stratum": "unseen",
468
+ "observed_coordinates": 28,
469
+ "remaining_coordinates": 28,
470
+ "records": 65,
471
+ "forecast_nll_per_remaining_dimension": 0.19157459161431303
472
+ },
473
+ {
474
+ "samples": 64,
475
+ "seed": 20852,
476
+ "model": "factorized_neural",
477
+ "stratum": "seen",
478
+ "observed_coordinates": 28,
479
+ "remaining_coordinates": 28,
480
+ "records": 339,
481
+ "forecast_nll_per_remaining_dimension": 0.17578733979705832
482
+ },
483
+ {
484
+ "samples": 64,
485
+ "seed": 20852,
486
+ "model": "joint_neural",
487
+ "stratum": "unseen",
488
+ "observed_coordinates": 0,
489
+ "remaining_coordinates": 56,
490
+ "records": 65,
491
+ "forecast_nll_per_remaining_dimension": 0.5628402915000806
492
+ },
493
+ {
494
+ "samples": 64,
495
+ "seed": 20852,
496
+ "model": "joint_neural",
497
+ "stratum": "seen",
498
+ "observed_coordinates": 0,
499
+ "remaining_coordinates": 56,
500
+ "records": 339,
501
+ "forecast_nll_per_remaining_dimension": 0.5534689289262176
502
+ },
503
+ {
504
+ "samples": 64,
505
+ "seed": 20852,
506
+ "model": "joint_neural",
507
+ "stratum": "unseen",
508
+ "observed_coordinates": 8,
509
+ "remaining_coordinates": 48,
510
+ "records": 65,
511
+ "forecast_nll_per_remaining_dimension": 0.502342196753467
512
+ },
513
+ {
514
+ "samples": 64,
515
+ "seed": 20852,
516
+ "model": "joint_neural",
517
+ "stratum": "seen",
518
+ "observed_coordinates": 8,
519
+ "remaining_coordinates": 48,
520
+ "records": 339,
521
+ "forecast_nll_per_remaining_dimension": 0.49870667612738495
522
+ },
523
+ {
524
+ "samples": 64,
525
+ "seed": 20852,
526
+ "model": "joint_neural",
527
+ "stratum": "unseen",
528
+ "observed_coordinates": 28,
529
+ "remaining_coordinates": 28,
530
+ "records": 65,
531
+ "forecast_nll_per_remaining_dimension": 0.3027475252330315
532
+ },
533
+ {
534
+ "samples": 64,
535
+ "seed": 20852,
536
+ "model": "joint_neural",
537
+ "stratum": "seen",
538
+ "observed_coordinates": 28,
539
+ "remaining_coordinates": 28,
540
+ "records": 339,
541
+ "forecast_nll_per_remaining_dimension": 0.29394876418305455
542
+ },
543
+ {
544
+ "samples": 64,
545
+ "seed": 20852,
546
+ "model": "joint_gaussian",
547
+ "stratum": "unseen",
548
+ "observed_coordinates": 0,
549
+ "remaining_coordinates": 56,
550
+ "records": 65,
551
+ "forecast_nll_per_remaining_dimension": 0.5494421683039526
552
+ },
553
+ {
554
+ "samples": 64,
555
+ "seed": 20852,
556
+ "model": "joint_gaussian",
557
+ "stratum": "seen",
558
+ "observed_coordinates": 0,
559
+ "remaining_coordinates": 56,
560
+ "records": 339,
561
+ "forecast_nll_per_remaining_dimension": 0.5451971920330484
562
+ },
563
+ {
564
+ "samples": 64,
565
+ "seed": 20852,
566
+ "model": "joint_gaussian",
567
+ "stratum": "unseen",
568
+ "observed_coordinates": 8,
569
+ "remaining_coordinates": 48,
570
+ "records": 65,
571
+ "forecast_nll_per_remaining_dimension": 0.49356456327098275
572
+ },
573
+ {
574
+ "samples": 64,
575
+ "seed": 20852,
576
+ "model": "joint_gaussian",
577
+ "stratum": "seen",
578
+ "observed_coordinates": 8,
579
+ "remaining_coordinates": 48,
580
+ "records": 339,
581
+ "forecast_nll_per_remaining_dimension": 0.4998280661105293
582
+ },
583
+ {
584
+ "samples": 64,
585
+ "seed": 20852,
586
+ "model": "joint_gaussian",
587
+ "stratum": "unseen",
588
+ "observed_coordinates": 28,
589
+ "remaining_coordinates": 28,
590
+ "records": 65,
591
+ "forecast_nll_per_remaining_dimension": 0.35506697899122786
592
+ },
593
+ {
594
+ "samples": 64,
595
+ "seed": 20852,
596
+ "model": "joint_gaussian",
597
+ "stratum": "seen",
598
+ "observed_coordinates": 28,
599
+ "remaining_coordinates": 28,
600
+ "records": 339,
601
+ "forecast_nll_per_remaining_dimension": 0.3548974304484435
602
+ },
603
+ {
604
+ "samples": 64,
605
+ "seed": 20853,
606
+ "model": "additive_full",
607
+ "stratum": "unseen",
608
+ "observed_coordinates": 0,
609
+ "remaining_coordinates": 56,
610
+ "records": 65,
611
+ "forecast_nll_per_remaining_dimension": 0.5704256157038452
612
+ },
613
+ {
614
+ "samples": 64,
615
+ "seed": 20853,
616
+ "model": "additive_full",
617
+ "stratum": "seen",
618
+ "observed_coordinates": 0,
619
+ "remaining_coordinates": 56,
620
+ "records": 339,
621
+ "forecast_nll_per_remaining_dimension": 0.5537161598862389
622
+ },
623
+ {
624
+ "samples": 64,
625
+ "seed": 20853,
626
+ "model": "additive_full",
627
+ "stratum": "unseen",
628
+ "observed_coordinates": 8,
629
+ "remaining_coordinates": 48,
630
+ "records": 65,
631
+ "forecast_nll_per_remaining_dimension": 0.4884359838135198
632
+ },
633
+ {
634
+ "samples": 64,
635
+ "seed": 20853,
636
+ "model": "additive_full",
637
+ "stratum": "seen",
638
+ "observed_coordinates": 8,
639
+ "remaining_coordinates": 48,
640
+ "records": 339,
641
+ "forecast_nll_per_remaining_dimension": 0.4914813157856711
642
+ },
643
+ {
644
+ "samples": 64,
645
+ "seed": 20853,
646
+ "model": "additive_full",
647
+ "stratum": "unseen",
648
+ "observed_coordinates": 28,
649
+ "remaining_coordinates": 28,
650
+ "records": 65,
651
+ "forecast_nll_per_remaining_dimension": 0.3690861485041636
652
+ },
653
+ {
654
+ "samples": 64,
655
+ "seed": 20853,
656
+ "model": "additive_full",
657
+ "stratum": "seen",
658
+ "observed_coordinates": 28,
659
+ "remaining_coordinates": 28,
660
+ "records": 339,
661
+ "forecast_nll_per_remaining_dimension": 0.3717139038004526
662
+ },
663
+ {
664
+ "samples": 64,
665
+ "seed": 20853,
666
+ "model": "additive_diag",
667
+ "stratum": "unseen",
668
+ "observed_coordinates": 0,
669
+ "remaining_coordinates": 56,
670
+ "records": 65,
671
+ "forecast_nll_per_remaining_dimension": 1.1273806175836993
672
+ },
673
+ {
674
+ "samples": 64,
675
+ "seed": 20853,
676
+ "model": "additive_diag",
677
+ "stratum": "seen",
678
+ "observed_coordinates": 0,
679
+ "remaining_coordinates": 56,
680
+ "records": 339,
681
+ "forecast_nll_per_remaining_dimension": 1.0313356069229929
682
+ },
683
+ {
684
+ "samples": 64,
685
+ "seed": 20853,
686
+ "model": "additive_diag",
687
+ "stratum": "unseen",
688
+ "observed_coordinates": 8,
689
+ "remaining_coordinates": 48,
690
+ "records": 65,
691
+ "forecast_nll_per_remaining_dimension": 1.0210015265538337
692
+ },
693
+ {
694
+ "samples": 64,
695
+ "seed": 20853,
696
+ "model": "additive_diag",
697
+ "stratum": "seen",
698
+ "observed_coordinates": 8,
699
+ "remaining_coordinates": 48,
700
+ "records": 339,
701
+ "forecast_nll_per_remaining_dimension": 0.9755538324717234
702
+ },
703
+ {
704
+ "samples": 64,
705
+ "seed": 20853,
706
+ "model": "additive_diag",
707
+ "stratum": "unseen",
708
+ "observed_coordinates": 28,
709
+ "remaining_coordinates": 28,
710
+ "records": 65,
711
+ "forecast_nll_per_remaining_dimension": 0.8747657855814529
712
+ },
713
+ {
714
+ "samples": 64,
715
+ "seed": 20853,
716
+ "model": "additive_diag",
717
+ "stratum": "seen",
718
+ "observed_coordinates": 28,
719
+ "remaining_coordinates": 28,
720
+ "records": 339,
721
+ "forecast_nll_per_remaining_dimension": 0.8709952402944175
722
+ },
723
+ {
724
+ "samples": 64,
725
+ "seed": 20853,
726
+ "model": "factorized_neural",
727
+ "stratum": "unseen",
728
+ "observed_coordinates": 0,
729
+ "remaining_coordinates": 56,
730
+ "records": 65,
731
+ "forecast_nll_per_remaining_dimension": 0.4597896813164183
732
+ },
733
+ {
734
+ "samples": 64,
735
+ "seed": 20853,
736
+ "model": "factorized_neural",
737
+ "stratum": "seen",
738
+ "observed_coordinates": 0,
739
+ "remaining_coordinates": 56,
740
+ "records": 339,
741
+ "forecast_nll_per_remaining_dimension": 0.44450240102390653
742
+ },
743
+ {
744
+ "samples": 64,
745
+ "seed": 20853,
746
+ "model": "factorized_neural",
747
+ "stratum": "unseen",
748
+ "observed_coordinates": 8,
749
+ "remaining_coordinates": 48,
750
+ "records": 65,
751
+ "forecast_nll_per_remaining_dimension": 0.33765717504146303
752
+ },
753
+ {
754
+ "samples": 64,
755
+ "seed": 20853,
756
+ "model": "factorized_neural",
757
+ "stratum": "seen",
758
+ "observed_coordinates": 8,
759
+ "remaining_coordinates": 48,
760
+ "records": 339,
761
+ "forecast_nll_per_remaining_dimension": 0.33058519074721704
762
+ },
763
+ {
764
+ "samples": 64,
765
+ "seed": 20853,
766
+ "model": "factorized_neural",
767
+ "stratum": "unseen",
768
+ "observed_coordinates": 28,
769
+ "remaining_coordinates": 28,
770
+ "records": 65,
771
+ "forecast_nll_per_remaining_dimension": -0.018818480023961026
772
+ },
773
+ {
774
+ "samples": 64,
775
+ "seed": 20853,
776
+ "model": "factorized_neural",
777
+ "stratum": "seen",
778
+ "observed_coordinates": 28,
779
+ "remaining_coordinates": 28,
780
+ "records": 339,
781
+ "forecast_nll_per_remaining_dimension": -0.033035717572371365
782
+ },
783
+ {
784
+ "samples": 64,
785
+ "seed": 20853,
786
+ "model": "joint_neural",
787
+ "stratum": "unseen",
788
+ "observed_coordinates": 0,
789
+ "remaining_coordinates": 56,
790
+ "records": 65,
791
+ "forecast_nll_per_remaining_dimension": 0.5364241216101618
792
+ },
793
+ {
794
+ "samples": 64,
795
+ "seed": 20853,
796
+ "model": "joint_neural",
797
+ "stratum": "seen",
798
+ "observed_coordinates": 0,
799
+ "remaining_coordinates": 56,
800
+ "records": 339,
801
+ "forecast_nll_per_remaining_dimension": 0.5224036981230613
802
+ },
803
+ {
804
+ "samples": 64,
805
+ "seed": 20853,
806
+ "model": "joint_neural",
807
+ "stratum": "unseen",
808
+ "observed_coordinates": 8,
809
+ "remaining_coordinates": 48,
810
+ "records": 65,
811
+ "forecast_nll_per_remaining_dimension": 0.4295827999137654
812
+ },
813
+ {
814
+ "samples": 64,
815
+ "seed": 20853,
816
+ "model": "joint_neural",
817
+ "stratum": "seen",
818
+ "observed_coordinates": 8,
819
+ "remaining_coordinates": 48,
820
+ "records": 339,
821
+ "forecast_nll_per_remaining_dimension": 0.4250680211739283
822
+ },
823
+ {
824
+ "samples": 64,
825
+ "seed": 20853,
826
+ "model": "joint_neural",
827
+ "stratum": "unseen",
828
+ "observed_coordinates": 28,
829
+ "remaining_coordinates": 28,
830
+ "records": 65,
831
+ "forecast_nll_per_remaining_dimension": 0.14400990368667058
832
+ },
833
+ {
834
+ "samples": 64,
835
+ "seed": 20853,
836
+ "model": "joint_neural",
837
+ "stratum": "seen",
838
+ "observed_coordinates": 28,
839
+ "remaining_coordinates": 28,
840
+ "records": 339,
841
+ "forecast_nll_per_remaining_dimension": 0.1336351412112061
842
+ },
843
+ {
844
+ "samples": 64,
845
+ "seed": 20853,
846
+ "model": "joint_gaussian",
847
+ "stratum": "unseen",
848
+ "observed_coordinates": 0,
849
+ "remaining_coordinates": 56,
850
+ "records": 65,
851
+ "forecast_nll_per_remaining_dimension": 0.5276462888856385
852
+ },
853
+ {
854
+ "samples": 64,
855
+ "seed": 20853,
856
+ "model": "joint_gaussian",
857
+ "stratum": "seen",
858
+ "observed_coordinates": 0,
859
+ "remaining_coordinates": 56,
860
+ "records": 339,
861
+ "forecast_nll_per_remaining_dimension": 0.5119187655445903
862
+ },
863
+ {
864
+ "samples": 64,
865
+ "seed": 20853,
866
+ "model": "joint_gaussian",
867
+ "stratum": "unseen",
868
+ "observed_coordinates": 8,
869
+ "remaining_coordinates": 48,
870
+ "records": 65,
871
+ "forecast_nll_per_remaining_dimension": 0.4389817364683426
872
+ },
873
+ {
874
+ "samples": 64,
875
+ "seed": 20853,
876
+ "model": "joint_gaussian",
877
+ "stratum": "seen",
878
+ "observed_coordinates": 8,
879
+ "remaining_coordinates": 48,
880
+ "records": 339,
881
+ "forecast_nll_per_remaining_dimension": 0.4419966726642397
882
+ },
883
+ {
884
+ "samples": 64,
885
+ "seed": 20853,
886
+ "model": "joint_gaussian",
887
+ "stratum": "unseen",
888
+ "observed_coordinates": 28,
889
+ "remaining_coordinates": 28,
890
+ "records": 65,
891
+ "forecast_nll_per_remaining_dimension": 0.27140773283955427
892
+ },
893
+ {
894
+ "samples": 64,
895
+ "seed": 20853,
896
+ "model": "joint_gaussian",
897
+ "stratum": "seen",
898
+ "observed_coordinates": 28,
899
+ "remaining_coordinates": 28,
900
+ "records": 339,
901
+ "forecast_nll_per_remaining_dimension": 0.27193780850353416
902
+ },
903
+ {
904
+ "samples": 128,
905
+ "seed": 20851,
906
+ "model": "additive_full",
907
+ "stratum": "unseen",
908
+ "observed_coordinates": 0,
909
+ "remaining_coordinates": 56,
910
+ "records": 65,
911
+ "forecast_nll_per_remaining_dimension": 0.48950039594021627
912
+ },
913
+ {
914
+ "samples": 128,
915
+ "seed": 20851,
916
+ "model": "additive_full",
917
+ "stratum": "seen",
918
+ "observed_coordinates": 0,
919
+ "remaining_coordinates": 56,
920
+ "records": 339,
921
+ "forecast_nll_per_remaining_dimension": 0.478921398442516
922
+ },
923
+ {
924
+ "samples": 128,
925
+ "seed": 20851,
926
+ "model": "additive_full",
927
+ "stratum": "unseen",
928
+ "observed_coordinates": 8,
929
+ "remaining_coordinates": 48,
930
+ "records": 65,
931
+ "forecast_nll_per_remaining_dimension": 0.4170756305774024
932
+ },
933
+ {
934
+ "samples": 128,
935
+ "seed": 20851,
936
+ "model": "additive_full",
937
+ "stratum": "seen",
938
+ "observed_coordinates": 8,
939
+ "remaining_coordinates": 48,
940
+ "records": 339,
941
+ "forecast_nll_per_remaining_dimension": 0.41841968024609
942
+ },
943
+ {
944
+ "samples": 128,
945
+ "seed": 20851,
946
+ "model": "additive_full",
947
+ "stratum": "unseen",
948
+ "observed_coordinates": 28,
949
+ "remaining_coordinates": 28,
950
+ "records": 65,
951
+ "forecast_nll_per_remaining_dimension": 0.2703191186119996
952
+ },
953
+ {
954
+ "samples": 128,
955
+ "seed": 20851,
956
+ "model": "additive_full",
957
+ "stratum": "seen",
958
+ "observed_coordinates": 28,
959
+ "remaining_coordinates": 28,
960
+ "records": 339,
961
+ "forecast_nll_per_remaining_dimension": 0.2660431128244684
962
+ },
963
+ {
964
+ "samples": 128,
965
+ "seed": 20851,
966
+ "model": "additive_diag",
967
+ "stratum": "unseen",
968
+ "observed_coordinates": 0,
969
+ "remaining_coordinates": 56,
970
+ "records": 65,
971
+ "forecast_nll_per_remaining_dimension": 1.0535688770417062
972
+ },
973
+ {
974
+ "samples": 128,
975
+ "seed": 20851,
976
+ "model": "additive_diag",
977
+ "stratum": "seen",
978
+ "observed_coordinates": 0,
979
+ "remaining_coordinates": 56,
980
+ "records": 339,
981
+ "forecast_nll_per_remaining_dimension": 0.9383546853703013
982
+ },
983
+ {
984
+ "samples": 128,
985
+ "seed": 20851,
986
+ "model": "additive_diag",
987
+ "stratum": "unseen",
988
+ "observed_coordinates": 8,
989
+ "remaining_coordinates": 48,
990
+ "records": 65,
991
+ "forecast_nll_per_remaining_dimension": 0.9883501067013684
992
+ },
993
+ {
994
+ "samples": 128,
995
+ "seed": 20851,
996
+ "model": "additive_diag",
997
+ "stratum": "seen",
998
+ "observed_coordinates": 8,
999
+ "remaining_coordinates": 48,
1000
+ "records": 339,
1001
+ "forecast_nll_per_remaining_dimension": 0.8961298320955366
1002
+ },
1003
+ {
1004
+ "samples": 128,
1005
+ "seed": 20851,
1006
+ "model": "additive_diag",
1007
+ "stratum": "unseen",
1008
+ "observed_coordinates": 28,
1009
+ "remaining_coordinates": 28,
1010
+ "records": 65,
1011
+ "forecast_nll_per_remaining_dimension": 0.893507116285307
1012
+ },
1013
+ {
1014
+ "samples": 128,
1015
+ "seed": 20851,
1016
+ "model": "additive_diag",
1017
+ "stratum": "seen",
1018
+ "observed_coordinates": 28,
1019
+ "remaining_coordinates": 28,
1020
+ "records": 339,
1021
+ "forecast_nll_per_remaining_dimension": 0.7881389404059385
1022
+ },
1023
+ {
1024
+ "samples": 128,
1025
+ "seed": 20851,
1026
+ "model": "factorized_neural",
1027
+ "stratum": "unseen",
1028
+ "observed_coordinates": 0,
1029
+ "remaining_coordinates": 56,
1030
+ "records": 65,
1031
+ "forecast_nll_per_remaining_dimension": 0.3791739796920747
1032
+ },
1033
+ {
1034
+ "samples": 128,
1035
+ "seed": 20851,
1036
+ "model": "factorized_neural",
1037
+ "stratum": "seen",
1038
+ "observed_coordinates": 0,
1039
+ "remaining_coordinates": 56,
1040
+ "records": 339,
1041
+ "forecast_nll_per_remaining_dimension": 0.36507694489804976
1042
+ },
1043
+ {
1044
+ "samples": 128,
1045
+ "seed": 20851,
1046
+ "model": "factorized_neural",
1047
+ "stratum": "unseen",
1048
+ "observed_coordinates": 8,
1049
+ "remaining_coordinates": 48,
1050
+ "records": 65,
1051
+ "forecast_nll_per_remaining_dimension": 0.28919680889651217
1052
+ },
1053
+ {
1054
+ "samples": 128,
1055
+ "seed": 20851,
1056
+ "model": "factorized_neural",
1057
+ "stratum": "seen",
1058
+ "observed_coordinates": 8,
1059
+ "remaining_coordinates": 48,
1060
+ "records": 339,
1061
+ "forecast_nll_per_remaining_dimension": 0.28017115466704495
1062
+ },
1063
+ {
1064
+ "samples": 128,
1065
+ "seed": 20851,
1066
+ "model": "factorized_neural",
1067
+ "stratum": "unseen",
1068
+ "observed_coordinates": 28,
1069
+ "remaining_coordinates": 28,
1070
+ "records": 65,
1071
+ "forecast_nll_per_remaining_dimension": 0.01442497192685894
1072
+ },
1073
+ {
1074
+ "samples": 128,
1075
+ "seed": 20851,
1076
+ "model": "factorized_neural",
1077
+ "stratum": "seen",
1078
+ "observed_coordinates": 28,
1079
+ "remaining_coordinates": 28,
1080
+ "records": 339,
1081
+ "forecast_nll_per_remaining_dimension": -0.006037562544335829
1082
+ },
1083
+ {
1084
+ "samples": 128,
1085
+ "seed": 20851,
1086
+ "model": "joint_neural",
1087
+ "stratum": "unseen",
1088
+ "observed_coordinates": 0,
1089
+ "remaining_coordinates": 56,
1090
+ "records": 65,
1091
+ "forecast_nll_per_remaining_dimension": 0.3993192592039324
1092
+ },
1093
+ {
1094
+ "samples": 128,
1095
+ "seed": 20851,
1096
+ "model": "joint_neural",
1097
+ "stratum": "seen",
1098
+ "observed_coordinates": 0,
1099
+ "remaining_coordinates": 56,
1100
+ "records": 339,
1101
+ "forecast_nll_per_remaining_dimension": 0.38419690536915946
1102
+ },
1103
+ {
1104
+ "samples": 128,
1105
+ "seed": 20851,
1106
+ "model": "joint_neural",
1107
+ "stratum": "unseen",
1108
+ "observed_coordinates": 8,
1109
+ "remaining_coordinates": 48,
1110
+ "records": 65,
1111
+ "forecast_nll_per_remaining_dimension": 0.3133013537172805
1112
+ },
1113
+ {
1114
+ "samples": 128,
1115
+ "seed": 20851,
1116
+ "model": "joint_neural",
1117
+ "stratum": "seen",
1118
+ "observed_coordinates": 8,
1119
+ "remaining_coordinates": 48,
1120
+ "records": 339,
1121
+ "forecast_nll_per_remaining_dimension": 0.3034748399094614
1122
+ },
1123
+ {
1124
+ "samples": 128,
1125
+ "seed": 20851,
1126
+ "model": "joint_neural",
1127
+ "stratum": "unseen",
1128
+ "observed_coordinates": 28,
1129
+ "remaining_coordinates": 28,
1130
+ "records": 65,
1131
+ "forecast_nll_per_remaining_dimension": 0.054119070617525904
1132
+ },
1133
+ {
1134
+ "samples": 128,
1135
+ "seed": 20851,
1136
+ "model": "joint_neural",
1137
+ "stratum": "seen",
1138
+ "observed_coordinates": 28,
1139
+ "remaining_coordinates": 28,
1140
+ "records": 339,
1141
+ "forecast_nll_per_remaining_dimension": 0.032122861019195925
1142
+ },
1143
+ {
1144
+ "samples": 128,
1145
+ "seed": 20851,
1146
+ "model": "joint_gaussian",
1147
+ "stratum": "unseen",
1148
+ "observed_coordinates": 0,
1149
+ "remaining_coordinates": 56,
1150
+ "records": 65,
1151
+ "forecast_nll_per_remaining_dimension": 0.43851553646057745
1152
+ },
1153
+ {
1154
+ "samples": 128,
1155
+ "seed": 20851,
1156
+ "model": "joint_gaussian",
1157
+ "stratum": "seen",
1158
+ "observed_coordinates": 0,
1159
+ "remaining_coordinates": 56,
1160
+ "records": 339,
1161
+ "forecast_nll_per_remaining_dimension": 0.4281615101503747
1162
+ },
1163
+ {
1164
+ "samples": 128,
1165
+ "seed": 20851,
1166
+ "model": "joint_gaussian",
1167
+ "stratum": "unseen",
1168
+ "observed_coordinates": 8,
1169
+ "remaining_coordinates": 48,
1170
+ "records": 65,
1171
+ "forecast_nll_per_remaining_dimension": 0.3624519562229542
1172
+ },
1173
+ {
1174
+ "samples": 128,
1175
+ "seed": 20851,
1176
+ "model": "joint_gaussian",
1177
+ "stratum": "seen",
1178
+ "observed_coordinates": 8,
1179
+ "remaining_coordinates": 48,
1180
+ "records": 339,
1181
+ "forecast_nll_per_remaining_dimension": 0.3614036445761413
1182
+ },
1183
+ {
1184
+ "samples": 128,
1185
+ "seed": 20851,
1186
+ "model": "joint_gaussian",
1187
+ "stratum": "unseen",
1188
+ "observed_coordinates": 28,
1189
+ "remaining_coordinates": 28,
1190
+ "records": 65,
1191
+ "forecast_nll_per_remaining_dimension": 0.17533608079559349
1192
+ },
1193
+ {
1194
+ "samples": 128,
1195
+ "seed": 20851,
1196
+ "model": "joint_gaussian",
1197
+ "stratum": "seen",
1198
+ "observed_coordinates": 28,
1199
+ "remaining_coordinates": 28,
1200
+ "records": 339,
1201
+ "forecast_nll_per_remaining_dimension": 0.1645598331830719
1202
+ },
1203
+ {
1204
+ "samples": 128,
1205
+ "seed": 20852,
1206
+ "model": "additive_full",
1207
+ "stratum": "unseen",
1208
+ "observed_coordinates": 0,
1209
+ "remaining_coordinates": 56,
1210
+ "records": 65,
1211
+ "forecast_nll_per_remaining_dimension": 0.4369235735447674
1212
+ },
1213
+ {
1214
+ "samples": 128,
1215
+ "seed": 20852,
1216
+ "model": "additive_full",
1217
+ "stratum": "seen",
1218
+ "observed_coordinates": 0,
1219
+ "remaining_coordinates": 56,
1220
+ "records": 339,
1221
+ "forecast_nll_per_remaining_dimension": 0.4302808247948633
1222
+ },
1223
+ {
1224
+ "samples": 128,
1225
+ "seed": 20852,
1226
+ "model": "additive_full",
1227
+ "stratum": "unseen",
1228
+ "observed_coordinates": 8,
1229
+ "remaining_coordinates": 48,
1230
+ "records": 65,
1231
+ "forecast_nll_per_remaining_dimension": 0.368969380130796
1232
+ },
1233
+ {
1234
+ "samples": 128,
1235
+ "seed": 20852,
1236
+ "model": "additive_full",
1237
+ "stratum": "seen",
1238
+ "observed_coordinates": 8,
1239
+ "remaining_coordinates": 48,
1240
+ "records": 339,
1241
+ "forecast_nll_per_remaining_dimension": 0.3739652479061342
1242
+ },
1243
+ {
1244
+ "samples": 128,
1245
+ "seed": 20852,
1246
+ "model": "additive_full",
1247
+ "stratum": "unseen",
1248
+ "observed_coordinates": 28,
1249
+ "remaining_coordinates": 28,
1250
+ "records": 65,
1251
+ "forecast_nll_per_remaining_dimension": 0.23307090925396362
1252
+ },
1253
+ {
1254
+ "samples": 128,
1255
+ "seed": 20852,
1256
+ "model": "additive_full",
1257
+ "stratum": "seen",
1258
+ "observed_coordinates": 28,
1259
+ "remaining_coordinates": 28,
1260
+ "records": 339,
1261
+ "forecast_nll_per_remaining_dimension": 0.23419369789039224
1262
+ },
1263
+ {
1264
+ "samples": 128,
1265
+ "seed": 20852,
1266
+ "model": "additive_diag",
1267
+ "stratum": "unseen",
1268
+ "observed_coordinates": 0,
1269
+ "remaining_coordinates": 56,
1270
+ "records": 65,
1271
+ "forecast_nll_per_remaining_dimension": 0.9099593843413494
1272
+ },
1273
+ {
1274
+ "samples": 128,
1275
+ "seed": 20852,
1276
+ "model": "additive_diag",
1277
+ "stratum": "seen",
1278
+ "observed_coordinates": 0,
1279
+ "remaining_coordinates": 56,
1280
+ "records": 339,
1281
+ "forecast_nll_per_remaining_dimension": 0.8506909930239959
1282
+ },
1283
+ {
1284
+ "samples": 128,
1285
+ "seed": 20852,
1286
+ "model": "additive_diag",
1287
+ "stratum": "unseen",
1288
+ "observed_coordinates": 8,
1289
+ "remaining_coordinates": 48,
1290
+ "records": 65,
1291
+ "forecast_nll_per_remaining_dimension": 0.8416501053866418
1292
+ },
1293
+ {
1294
+ "samples": 128,
1295
+ "seed": 20852,
1296
+ "model": "additive_diag",
1297
+ "stratum": "seen",
1298
+ "observed_coordinates": 8,
1299
+ "remaining_coordinates": 48,
1300
+ "records": 339,
1301
+ "forecast_nll_per_remaining_dimension": 0.8114590572681132
1302
+ },
1303
+ {
1304
+ "samples": 128,
1305
+ "seed": 20852,
1306
+ "model": "additive_diag",
1307
+ "stratum": "unseen",
1308
+ "observed_coordinates": 28,
1309
+ "remaining_coordinates": 28,
1310
+ "records": 65,
1311
+ "forecast_nll_per_remaining_dimension": 0.6911617414076279
1312
+ },
1313
+ {
1314
+ "samples": 128,
1315
+ "seed": 20852,
1316
+ "model": "additive_diag",
1317
+ "stratum": "seen",
1318
+ "observed_coordinates": 28,
1319
+ "remaining_coordinates": 28,
1320
+ "records": 339,
1321
+ "forecast_nll_per_remaining_dimension": 0.7074443571547163
1322
+ },
1323
+ {
1324
+ "samples": 128,
1325
+ "seed": 20852,
1326
+ "model": "factorized_neural",
1327
+ "stratum": "unseen",
1328
+ "observed_coordinates": 0,
1329
+ "remaining_coordinates": 56,
1330
+ "records": 65,
1331
+ "forecast_nll_per_remaining_dimension": 0.3153330205819093
1332
+ },
1333
+ {
1334
+ "samples": 128,
1335
+ "seed": 20852,
1336
+ "model": "factorized_neural",
1337
+ "stratum": "seen",
1338
+ "observed_coordinates": 0,
1339
+ "remaining_coordinates": 56,
1340
+ "records": 339,
1341
+ "forecast_nll_per_remaining_dimension": 0.2913443706095867
1342
+ },
1343
+ {
1344
+ "samples": 128,
1345
+ "seed": 20852,
1346
+ "model": "factorized_neural",
1347
+ "stratum": "unseen",
1348
+ "observed_coordinates": 8,
1349
+ "remaining_coordinates": 48,
1350
+ "records": 65,
1351
+ "forecast_nll_per_remaining_dimension": 0.2330889144168917
1352
+ },
1353
+ {
1354
+ "samples": 128,
1355
+ "seed": 20852,
1356
+ "model": "factorized_neural",
1357
+ "stratum": "seen",
1358
+ "observed_coordinates": 8,
1359
+ "remaining_coordinates": 48,
1360
+ "records": 339,
1361
+ "forecast_nll_per_remaining_dimension": 0.2136504781028952
1362
+ },
1363
+ {
1364
+ "samples": 128,
1365
+ "seed": 20852,
1366
+ "model": "factorized_neural",
1367
+ "stratum": "unseen",
1368
+ "observed_coordinates": 28,
1369
+ "remaining_coordinates": 28,
1370
+ "records": 65,
1371
+ "forecast_nll_per_remaining_dimension": -0.002574211280722225
1372
+ },
1373
+ {
1374
+ "samples": 128,
1375
+ "seed": 20852,
1376
+ "model": "factorized_neural",
1377
+ "stratum": "seen",
1378
+ "observed_coordinates": 28,
1379
+ "remaining_coordinates": 28,
1380
+ "records": 339,
1381
+ "forecast_nll_per_remaining_dimension": -0.03919904974260129
1382
+ },
1383
+ {
1384
+ "samples": 128,
1385
+ "seed": 20852,
1386
+ "model": "joint_neural",
1387
+ "stratum": "unseen",
1388
+ "observed_coordinates": 0,
1389
+ "remaining_coordinates": 56,
1390
+ "records": 65,
1391
+ "forecast_nll_per_remaining_dimension": 0.3603470718451079
1392
+ },
1393
+ {
1394
+ "samples": 128,
1395
+ "seed": 20852,
1396
+ "model": "joint_neural",
1397
+ "stratum": "seen",
1398
+ "observed_coordinates": 0,
1399
+ "remaining_coordinates": 56,
1400
+ "records": 339,
1401
+ "forecast_nll_per_remaining_dimension": 0.33623607500582287
1402
+ },
1403
+ {
1404
+ "samples": 128,
1405
+ "seed": 20852,
1406
+ "model": "joint_neural",
1407
+ "stratum": "unseen",
1408
+ "observed_coordinates": 8,
1409
+ "remaining_coordinates": 48,
1410
+ "records": 65,
1411
+ "forecast_nll_per_remaining_dimension": 0.2885899354588873
1412
+ },
1413
+ {
1414
+ "samples": 128,
1415
+ "seed": 20852,
1416
+ "model": "joint_neural",
1417
+ "stratum": "seen",
1418
+ "observed_coordinates": 8,
1419
+ "remaining_coordinates": 48,
1420
+ "records": 339,
1421
+ "forecast_nll_per_remaining_dimension": 0.26995143965296126
1422
+ },
1423
+ {
1424
+ "samples": 128,
1425
+ "seed": 20852,
1426
+ "model": "joint_neural",
1427
+ "stratum": "unseen",
1428
+ "observed_coordinates": 28,
1429
+ "remaining_coordinates": 28,
1430
+ "records": 65,
1431
+ "forecast_nll_per_remaining_dimension": 0.09332468659065243
1432
+ },
1433
+ {
1434
+ "samples": 128,
1435
+ "seed": 20852,
1436
+ "model": "joint_neural",
1437
+ "stratum": "seen",
1438
+ "observed_coordinates": 28,
1439
+ "remaining_coordinates": 28,
1440
+ "records": 339,
1441
+ "forecast_nll_per_remaining_dimension": 0.058101489658363495
1442
+ },
1443
+ {
1444
+ "samples": 128,
1445
+ "seed": 20852,
1446
+ "model": "joint_gaussian",
1447
+ "stratum": "unseen",
1448
+ "observed_coordinates": 0,
1449
+ "remaining_coordinates": 56,
1450
+ "records": 65,
1451
+ "forecast_nll_per_remaining_dimension": 0.3747684000670196
1452
+ },
1453
+ {
1454
+ "samples": 128,
1455
+ "seed": 20852,
1456
+ "model": "joint_gaussian",
1457
+ "stratum": "seen",
1458
+ "observed_coordinates": 0,
1459
+ "remaining_coordinates": 56,
1460
+ "records": 339,
1461
+ "forecast_nll_per_remaining_dimension": 0.3592348486895685
1462
+ },
1463
+ {
1464
+ "samples": 128,
1465
+ "seed": 20852,
1466
+ "model": "joint_gaussian",
1467
+ "stratum": "unseen",
1468
+ "observed_coordinates": 8,
1469
+ "remaining_coordinates": 48,
1470
+ "records": 65,
1471
+ "forecast_nll_per_remaining_dimension": 0.3096444105604308
1472
+ },
1473
+ {
1474
+ "samples": 128,
1475
+ "seed": 20852,
1476
+ "model": "joint_gaussian",
1477
+ "stratum": "seen",
1478
+ "observed_coordinates": 8,
1479
+ "remaining_coordinates": 48,
1480
+ "records": 339,
1481
+ "forecast_nll_per_remaining_dimension": 0.3034617999119028
1482
+ },
1483
+ {
1484
+ "samples": 128,
1485
+ "seed": 20852,
1486
+ "model": "joint_gaussian",
1487
+ "stratum": "unseen",
1488
+ "observed_coordinates": 28,
1489
+ "remaining_coordinates": 28,
1490
+ "records": 65,
1491
+ "forecast_nll_per_remaining_dimension": 0.15325446788177188
1492
+ },
1493
+ {
1494
+ "samples": 128,
1495
+ "seed": 20852,
1496
+ "model": "joint_gaussian",
1497
+ "stratum": "seen",
1498
+ "observed_coordinates": 28,
1499
+ "remaining_coordinates": 28,
1500
+ "records": 339,
1501
+ "forecast_nll_per_remaining_dimension": 0.13780706857656624
1502
+ },
1503
+ {
1504
+ "samples": 128,
1505
+ "seed": 20853,
1506
+ "model": "additive_full",
1507
+ "stratum": "unseen",
1508
+ "observed_coordinates": 0,
1509
+ "remaining_coordinates": 56,
1510
+ "records": 65,
1511
+ "forecast_nll_per_remaining_dimension": 0.4992420253096935
1512
+ },
1513
+ {
1514
+ "samples": 128,
1515
+ "seed": 20853,
1516
+ "model": "additive_full",
1517
+ "stratum": "seen",
1518
+ "observed_coordinates": 0,
1519
+ "remaining_coordinates": 56,
1520
+ "records": 339,
1521
+ "forecast_nll_per_remaining_dimension": 0.47220824616860746
1522
+ },
1523
+ {
1524
+ "samples": 128,
1525
+ "seed": 20853,
1526
+ "model": "additive_full",
1527
+ "stratum": "unseen",
1528
+ "observed_coordinates": 8,
1529
+ "remaining_coordinates": 48,
1530
+ "records": 65,
1531
+ "forecast_nll_per_remaining_dimension": 0.40160248660104014
1532
+ },
1533
+ {
1534
+ "samples": 128,
1535
+ "seed": 20853,
1536
+ "model": "additive_full",
1537
+ "stratum": "seen",
1538
+ "observed_coordinates": 8,
1539
+ "remaining_coordinates": 48,
1540
+ "records": 339,
1541
+ "forecast_nll_per_remaining_dimension": 0.3951007171013775
1542
+ },
1543
+ {
1544
+ "samples": 128,
1545
+ "seed": 20853,
1546
+ "model": "additive_full",
1547
+ "stratum": "unseen",
1548
+ "observed_coordinates": 28,
1549
+ "remaining_coordinates": 28,
1550
+ "records": 65,
1551
+ "forecast_nll_per_remaining_dimension": 0.2394866278290625
1552
+ },
1553
+ {
1554
+ "samples": 128,
1555
+ "seed": 20853,
1556
+ "model": "additive_full",
1557
+ "stratum": "seen",
1558
+ "observed_coordinates": 28,
1559
+ "remaining_coordinates": 28,
1560
+ "records": 339,
1561
+ "forecast_nll_per_remaining_dimension": 0.22896130967689032
1562
+ },
1563
+ {
1564
+ "samples": 128,
1565
+ "seed": 20853,
1566
+ "model": "additive_diag",
1567
+ "stratum": "unseen",
1568
+ "observed_coordinates": 0,
1569
+ "remaining_coordinates": 56,
1570
+ "records": 65,
1571
+ "forecast_nll_per_remaining_dimension": 1.1334353166190172
1572
+ },
1573
+ {
1574
+ "samples": 128,
1575
+ "seed": 20853,
1576
+ "model": "additive_diag",
1577
+ "stratum": "seen",
1578
+ "observed_coordinates": 0,
1579
+ "remaining_coordinates": 56,
1580
+ "records": 339,
1581
+ "forecast_nll_per_remaining_dimension": 0.9559373817120549
1582
+ },
1583
+ {
1584
+ "samples": 128,
1585
+ "seed": 20853,
1586
+ "model": "additive_diag",
1587
+ "stratum": "unseen",
1588
+ "observed_coordinates": 8,
1589
+ "remaining_coordinates": 48,
1590
+ "records": 65,
1591
+ "forecast_nll_per_remaining_dimension": 1.0117658277900303
1592
+ },
1593
+ {
1594
+ "samples": 128,
1595
+ "seed": 20853,
1596
+ "model": "additive_diag",
1597
+ "stratum": "seen",
1598
+ "observed_coordinates": 8,
1599
+ "remaining_coordinates": 48,
1600
+ "records": 339,
1601
+ "forecast_nll_per_remaining_dimension": 0.8810679057700632
1602
+ },
1603
+ {
1604
+ "samples": 128,
1605
+ "seed": 20853,
1606
+ "model": "additive_diag",
1607
+ "stratum": "unseen",
1608
+ "observed_coordinates": 28,
1609
+ "remaining_coordinates": 28,
1610
+ "records": 65,
1611
+ "forecast_nll_per_remaining_dimension": 0.8786702250014987
1612
+ },
1613
+ {
1614
+ "samples": 128,
1615
+ "seed": 20853,
1616
+ "model": "additive_diag",
1617
+ "stratum": "seen",
1618
+ "observed_coordinates": 28,
1619
+ "remaining_coordinates": 28,
1620
+ "records": 339,
1621
+ "forecast_nll_per_remaining_dimension": 0.7362512572648624
1622
+ },
1623
+ {
1624
+ "samples": 128,
1625
+ "seed": 20853,
1626
+ "model": "factorized_neural",
1627
+ "stratum": "unseen",
1628
+ "observed_coordinates": 0,
1629
+ "remaining_coordinates": 56,
1630
+ "records": 65,
1631
+ "forecast_nll_per_remaining_dimension": 0.415158507239599
1632
+ },
1633
+ {
1634
+ "samples": 128,
1635
+ "seed": 20853,
1636
+ "model": "factorized_neural",
1637
+ "stratum": "seen",
1638
+ "observed_coordinates": 0,
1639
+ "remaining_coordinates": 56,
1640
+ "records": 339,
1641
+ "forecast_nll_per_remaining_dimension": 0.39045017532560866
1642
+ },
1643
+ {
1644
+ "samples": 128,
1645
+ "seed": 20853,
1646
+ "model": "factorized_neural",
1647
+ "stratum": "unseen",
1648
+ "observed_coordinates": 8,
1649
+ "remaining_coordinates": 48,
1650
+ "records": 65,
1651
+ "forecast_nll_per_remaining_dimension": 0.30015140077254115
1652
+ },
1653
+ {
1654
+ "samples": 128,
1655
+ "seed": 20853,
1656
+ "model": "factorized_neural",
1657
+ "stratum": "seen",
1658
+ "observed_coordinates": 8,
1659
+ "remaining_coordinates": 48,
1660
+ "records": 339,
1661
+ "forecast_nll_per_remaining_dimension": 0.28740912419886755
1662
+ },
1663
+ {
1664
+ "samples": 128,
1665
+ "seed": 20853,
1666
+ "model": "factorized_neural",
1667
+ "stratum": "unseen",
1668
+ "observed_coordinates": 28,
1669
+ "remaining_coordinates": 28,
1670
+ "records": 65,
1671
+ "forecast_nll_per_remaining_dimension": 0.017067529437733688
1672
+ },
1673
+ {
1674
+ "samples": 128,
1675
+ "seed": 20853,
1676
+ "model": "factorized_neural",
1677
+ "stratum": "seen",
1678
+ "observed_coordinates": 28,
1679
+ "remaining_coordinates": 28,
1680
+ "records": 339,
1681
+ "forecast_nll_per_remaining_dimension": -0.004143085640689043
1682
+ },
1683
+ {
1684
+ "samples": 128,
1685
+ "seed": 20853,
1686
+ "model": "joint_neural",
1687
+ "stratum": "unseen",
1688
+ "observed_coordinates": 0,
1689
+ "remaining_coordinates": 56,
1690
+ "records": 65,
1691
+ "forecast_nll_per_remaining_dimension": 0.4306837685521369
1692
+ },
1693
+ {
1694
+ "samples": 128,
1695
+ "seed": 20853,
1696
+ "model": "joint_neural",
1697
+ "stratum": "seen",
1698
+ "observed_coordinates": 0,
1699
+ "remaining_coordinates": 56,
1700
+ "records": 339,
1701
+ "forecast_nll_per_remaining_dimension": 0.39732283614078673
1702
+ },
1703
+ {
1704
+ "samples": 128,
1705
+ "seed": 20853,
1706
+ "model": "joint_neural",
1707
+ "stratum": "unseen",
1708
+ "observed_coordinates": 8,
1709
+ "remaining_coordinates": 48,
1710
+ "records": 65,
1711
+ "forecast_nll_per_remaining_dimension": 0.3191750905914261
1712
+ },
1713
+ {
1714
+ "samples": 128,
1715
+ "seed": 20853,
1716
+ "model": "joint_neural",
1717
+ "stratum": "seen",
1718
+ "observed_coordinates": 8,
1719
+ "remaining_coordinates": 48,
1720
+ "records": 339,
1721
+ "forecast_nll_per_remaining_dimension": 0.29722289664793394
1722
+ },
1723
+ {
1724
+ "samples": 128,
1725
+ "seed": 20853,
1726
+ "model": "joint_neural",
1727
+ "stratum": "unseen",
1728
+ "observed_coordinates": 28,
1729
+ "remaining_coordinates": 28,
1730
+ "records": 65,
1731
+ "forecast_nll_per_remaining_dimension": 0.04808951878474744
1732
+ },
1733
+ {
1734
+ "samples": 128,
1735
+ "seed": 20853,
1736
+ "model": "joint_neural",
1737
+ "stratum": "seen",
1738
+ "observed_coordinates": 28,
1739
+ "remaining_coordinates": 28,
1740
+ "records": 339,
1741
+ "forecast_nll_per_remaining_dimension": 0.011713636329412068
1742
+ },
1743
+ {
1744
+ "samples": 128,
1745
+ "seed": 20853,
1746
+ "model": "joint_gaussian",
1747
+ "stratum": "unseen",
1748
+ "observed_coordinates": 0,
1749
+ "remaining_coordinates": 56,
1750
+ "records": 65,
1751
+ "forecast_nll_per_remaining_dimension": 0.4565969853344243
1752
+ },
1753
+ {
1754
+ "samples": 128,
1755
+ "seed": 20853,
1756
+ "model": "joint_gaussian",
1757
+ "stratum": "seen",
1758
+ "observed_coordinates": 0,
1759
+ "remaining_coordinates": 56,
1760
+ "records": 339,
1761
+ "forecast_nll_per_remaining_dimension": 0.42981770158257643
1762
+ },
1763
+ {
1764
+ "samples": 128,
1765
+ "seed": 20853,
1766
+ "model": "joint_gaussian",
1767
+ "stratum": "unseen",
1768
+ "observed_coordinates": 8,
1769
+ "remaining_coordinates": 48,
1770
+ "records": 65,
1771
+ "forecast_nll_per_remaining_dimension": 0.3546200654494273
1772
+ },
1773
+ {
1774
+ "samples": 128,
1775
+ "seed": 20853,
1776
+ "model": "joint_gaussian",
1777
+ "stratum": "seen",
1778
+ "observed_coordinates": 8,
1779
+ "remaining_coordinates": 48,
1780
+ "records": 339,
1781
+ "forecast_nll_per_remaining_dimension": 0.3469378761021978
1782
+ },
1783
+ {
1784
+ "samples": 128,
1785
+ "seed": 20853,
1786
+ "model": "joint_gaussian",
1787
+ "stratum": "unseen",
1788
+ "observed_coordinates": 28,
1789
+ "remaining_coordinates": 28,
1790
+ "records": 65,
1791
+ "forecast_nll_per_remaining_dimension": 0.15524344209853103
1792
+ },
1793
+ {
1794
+ "samples": 128,
1795
+ "seed": 20853,
1796
+ "model": "joint_gaussian",
1797
+ "stratum": "seen",
1798
+ "observed_coordinates": 28,
1799
+ "remaining_coordinates": 28,
1800
+ "records": 339,
1801
+ "forecast_nll_per_remaining_dimension": 0.14145341114575885
1802
+ }
1803
+ ],
1804
+ "seconds": 0.3921560939925257,
1805
+ "post_evaluation_mechanistic_analysis": true,
1806
+ "fixed_prefix_masks_never_used_to_select_or_fit": true,
1807
+ "withheld_measurements_accessed_only_by_offline_scoring": true,
1808
+ "units": "source-prefix standardized evidence density, nats per remaining coordinate"
1809
+ }
provenance/temperature/incomplete_cohort_evaluation.py ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Secondary fixed-model evaluation after source-cohort gate failure.
2
+ No architecture, feature, fitting prefix, held day, model threshold or query rule
3
+ is changed. The invalid source day was excluded by the ORIGINAL reader rule.
4
+ The complete13-day gate remains FAIL; no algorithm promotion is possible.
5
+ """
6
+ import json,time,torch
7
+ from unionmind.day_transfer import ROOT,PROTOCOL,MODELS,STRONG,verify,write,fit_one,predict,metrics,factorial,acquisition
8
+
9
+ def run(device='cuda:1'):
10
+ verify();torch.cuda.set_device(device);torch.backends.cuda.matmul.allow_tf32=False
11
+ config=json.loads(PROTOCOL.read_text());start=time.perf_counter();prep=json.loads((ROOT/'preparation.json').read_text())
12
+ write('incomplete_cohort_notice.json',dict(complete_cohort=prep['eligible'],promotion=False,source_day_2_excluded_by_original_validity_rule=True,unchanged_models_and_held_days=True))
13
+ p=torch.load(ROOT/'prepared.pt',map_location=device,weights_only=True);pool=torch.where(p['source'])[0];held=torch.where(p['target'])[0]
14
+ sizes=[10,3];grid=factorial.hypotheses(sizes,device);true=factorial.joint_index(p['h'][held],sizes);results=[]
15
+ for n in config['sample_budgets']:
16
+ for seed in config['seeds']:
17
+ rng=torch.Generator(device=device).manual_seed(seed);ids=pool[torch.randperm(len(pool),device=device,generator=rng)[:n]]
18
+ t=time.perf_counter();mean=p['x'][ids].mean(0);scale=p['x'][ids].std(0).clamp_min(1e-6)
19
+ x=(p['x'][ids]-mean)/scale;target=(p['x'][held]-mean)/scale;torch.cuda.synchronize();normalization=time.perf_counter()-t
20
+ supported=torch.ones(len(held),device=device,dtype=torch.bool)
21
+ for f,k in enumerate(sizes):supported &= torch.bincount(p['h'][ids,f],minlength=k)[p['h'][held,f]]>0
22
+ directory=ROOT/'models'/f'n{n}_s{seed}';directory.mkdir(parents=True,exist_ok=True)
23
+ for name in MODELS:
24
+ torch.cuda.synchronize();t=time.perf_counter();model=fit_one(x,p['h'][ids],sizes,name,seed)
25
+ torch.cuda.synchronize();fit_seconds=time.perf_counter()-t;t=time.perf_counter()
26
+ with torch.no_grad():prob=predict(model,target,range(x.shape[-1]));v=metrics(prob,true,grid,p['h'][held])
27
+ torch.cuda.synchronize();prediction=time.perf_counter()-t
28
+ checkpoint=directory/f'{name}.pt';torch.save(dict(model=model,mean=mean,scale=scale,fit_ids=ids),checkpoint)
29
+ row=dict(samples=n,seed=seed,model=name,held_rows=held,days=p['days'][held],unseen=p['unseen'][held],
30
+ supported_levels=supported,metrics=v,normalization_seconds=normalization,fit_seconds=fit_seconds,
31
+ prediction_seconds=prediction,neural_parameters=model.get('neural_parameters',0),
32
+ checkpoint_bytes=checkpoint.stat().st_size,adaptation_fits=model.get('adaptation_fits',1))
33
+ results.append(row)
34
+ print(json.dumps(dict(samples=n,seed=seed,model=name,top3_unseen=float(v['top3'][row['unseen']].mean()))),flush=True)
35
+ torch.save(dict(results=results),ROOT/'results.pt');write('execution.json',dict(seconds=time.perf_counter()-start,
36
+ peak_gpu_bytes=torch.cuda.max_memory_allocated(device),device=device,cells=len(results),backbone_executions=0,
37
+ timings=[{k:r[k] for k in ('samples','seed','model','normalization_seconds','fit_seconds','prediction_seconds','neural_parameters','checkpoint_bytes','adaptation_fits')} for r in results]))
38
+ verify()
39
+
40
+ @torch.no_grad()
41
+ def analyze(device='cuda:2'):
42
+ verify();torch.cuda.set_device(device);config=json.loads(PROTOCOL.read_text());start=time.perf_counter()
43
+ results=torch.load(ROOT/'results.pt',map_location=device,weights_only=True)['results'];rng=torch.Generator(device=device).manual_seed(20891)
44
+ summary=[];comparisons=[];gates=[]
45
+ for n in config['sample_budgets']:
46
+ records=[r for r in results if r['samples']==n];template=records[0];mask=template['unseen'];day=template['days'][mask]
47
+ matrix=torch.stack([torch.stack([r['metrics']['top3'] for r in records if r['model']==name]).mean(0)[mask] for name in MODELS],-1)
48
+ points=[];boot=[]
49
+ for d in range(9,13):
50
+ a=matrix[day==d];points.append(a.mean(0));sample=torch.randint(len(a),(20000,len(a)),device=device,generator=rng);boot.append(a[sample].mean(1))
51
+ day_points=torch.stack(points);point=day_points.mean(0);conditional_boot=torch.stack(boot).mean(0)
52
+ sampled_days=torch.randint(4,(20000,4),device=device,generator=rng)
53
+ day_boot=day_points[sampled_days].mean(1)
54
+ for name in MODELS:
55
+ rows=[r for r in records if r['model']==name];assert len(rows)==3
56
+ for stratum,include in (('unseen',mask),('seen',~mask)):
57
+ for d in (None,9,10,11,12):
58
+ m=include if d is None else include&(template['days']==d)
59
+ summary.append(dict(samples=n,model=name,stratum=stratum,day=d,records=int(m.sum()),
60
+ **{f:float(torch.stack([r['metrics'][f][m].mean() for r in rows]).mean()) for f in ('top1','top3','nll','confidence','top3_mass')}))
61
+ for support in (False,True):
62
+ count=correct=0
63
+ for r in rows:
64
+ m=r['unseen']&(r['supported_levels']==support);count+=int(m.sum());correct+=float(r['metrics']['top3'][m].sum())
65
+ summary.append(dict(samples=n,model=name,stratum='supported' if support else 'novel_level',records_across_seeds=count,top3=correct/count if count else None))
66
+ for threshold in (.5,.7,.9):
67
+ accepted=correct=0
68
+ for r in rows:
69
+ m=r['unseen']&(r['metrics']['confidence']>=threshold);accepted+=int(m.sum());correct+=float(r['metrics']['top1'][m].sum())
70
+ summary.append(dict(samples=n,model=name,stratum='selective_unseen',threshold=threshold,
71
+ coverage=accepted/(3*int(mask.sum())),accuracy=correct/accepted if accepted else None))
72
+ for candidate in config['candidates']:
73
+ ci=MODELS.index(candidate);rows=[];alpha=.05/(2*40)
74
+ for ri,reference in enumerate(MODELS):
75
+ if ci==ri:continue
76
+ diff=conditional_boot[:,ci]-conditional_boot[:,ri];lo,hi=torch.quantile(diff,diff.new_tensor([alpha,1-alpha]))
77
+ dd=day_boot[:,ci]-day_boot[:,ri];dlo,dhi=torch.quantile(dd,dd.new_tensor([alpha,1-alpha]))
78
+ row=dict(samples=n,candidate=candidate,control=reference,gain=float(point[ci]-point[ri]),
79
+ fixed_day_simultaneous_interval=[float(lo),float(hi)],day_cluster_simultaneous_interval=[float(dlo),float(dhi)],
80
+ day_gains=(day_points[:,ci]-day_points[:,ri]).tolist())
81
+ rows.append(row);comparisons.append(row)
82
+ strong=[r for r in rows if r['control'] in STRONG]
83
+ substantial=all(r['gain']>=.10 and r['fixed_day_simultaneous_interval'][0]>0 and r['day_cluster_simultaneous_interval'][0]>0 and min(r['day_gains'])>0 for r in strong)
84
+ best=all(r['fixed_day_simultaneous_interval'][0]>0 and r['day_cluster_simultaneous_interval'][0]>0 for r in rows)
85
+ learned=next((r['fixed_day_simultaneous_interval'][0]>0 and r['day_cluster_simultaneous_interval'][0]>0 for r in rows if r['control']=='additive_full'),False) if candidate=='factorized_neural' else None
86
+ gates.append(dict(samples=n,candidate=candidate,top3=float(point[ci]),substantial_gain=substantial,strongest_available=best,
87
+ learning_specific=learned,promotion=False,
88
+ strongest_observed_model=MODELS[int(point.argmax())],strongest_top3=float(point.max())))
89
+ write('summary.json',dict(rows=summary,comparisons=comparisons,gates=gates,all_thirteen_days_audited=True,complete_source_cohort=False,valid_source_days=8,all_four_target_days_retained=True,secondary_incomplete_cohort_evaluation=True,
90
+ confirmation_days=[9,10,11,12],one_physical_platform=True,draws=20000,comparison_family=40,
91
+ original_dynamic_mixture_source_gate_remains_failed=True))
92
+ write('analysis_execution.json',dict(seconds=time.perf_counter()-start,device=device));verify();print(json.dumps(gates),flush=True)
93
+
94
+ run()
95
+ analyze()
96
+ acquisition()
provenance/temperature/incomplete_cohort_evaluation_lock.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "sha256": "9d613b1c831a97c63fb3ac1eaafbd8763ced3ea4d7443635ffbe17722399224c",
3
+ "before_any_model_fit_or_prediction": true,
4
+ "source_cohort_gate": "FAIL and unchanged",
5
+ "reason": "one source-day clock is nonmonotonic; fixed remaining source pool can still be compared on all untouched target days",
6
+ "model_algorithms_unchanged": true
7
+ }
provenance/temperature/lock.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "/home/nurgaly/experiment/unionmind/unionmind/day_transfer.py": "af750c0f6f1c19923eb3e90acf9edaa56a998d9ae460b2d18002edfc0e33d774",
4
+ "configs/day_transfer_protocol_01.json": "a8a46886819f182a89e3ba00bb0cdc0be8b28d5988ce5a6bd8c7f3281e5a2f9b",
5
+ "unionmind/continuous_decision/classifiers.py": "ebb1bfcd9d72469368ffc14cbb847faeb53a69d19abc9b684546f3791d61e20b",
6
+ "unionmind/continuous_decision/__init__.py": "3a95b4d2e04edd607dbf129eb72b173f8addbc66086484720edea2b0e4e09d76",
7
+ "unionmind/continuous_decision/factor_acquisition.py": "771f560388ba9a24adadd7fd84d5bb90ff5f52e16ffc92e121a0aa4ba0a91b51",
8
+ "unionmind/continuous_decision/metric.py": "74919a6b6e5142da90cf63f3a154426ee195c699c744e34ce9207253544b2023",
9
+ "unionmind/continuous_decision/mixture.py": "2d25dfbc9df8cb9b471c247c88913fc9f91bc9aa6bc5c8d8e4e412fd7965612e",
10
+ "unionmind/continuous_decision/factor_study.py": "beb76b2cc09aaa2b796e8bf69f0235fb9b0c33823ce1e33ea8170d56b60ba109",
11
+ "unionmind/continuous_decision/learned_controls.py": "81e3419c8f20422ed495fab8b43fa264dceb8503d1a3ad71cdc70b4c32b0d6f3",
12
+ "unionmind/continuous_decision/acquisition_analysis.py": "75bd00b2c5374c4ad11588285a92d589509e8798b9c432bd5c0484aa54dcd6a1",
13
+ "unionmind/continuous_decision/factor_analysis.py": "0fb267e3c0572db48f4a2ca41fc4777ec082039c8312f66a75005355837c3fb8",
14
+ "unionmind/continuous_decision/interval_value.py": "cd3c0158a377c97fa9f4dc2f56c2a9597840a68bb07d37782b42265f727e4fe9",
15
+ "unionmind/continuous_decision/factorial.py": "54241b65b32042937346369414d750fe58042f0eae20f25c391a74a480e065f0",
16
+ "unionmind/continuous_decision/source.py": "1c011b7608637c73652224573caf152cefc3297c2f94599b8f88816ded829601",
17
+ "unionmind/continuous_decision/gaussian.py": "588a3e5f39a79fe26b959b7b2a62c6700bdd242e03a6c2adbd674f8d6537a2e5",
18
+ "tests/test_continuous_decision.py": "b5e0f4e887f8483b41b56e179b10bf9349d5a18367eb47f5634c98a8ffb023ce"
19
+ },
20
+ "before_archive_or_numerical_access": true,
21
+ "timestamp_utc": "2026-10-20T13:01:49Z",
22
+ "previous_strict_clock_study_not_amended": true,
23
+ "unchanged_algorithms_and_hyperparameters": true
24
+ }